Time to evolve: the applicability of pain phenotyping in manual therapy

Physical therapists use a number of tools for pain management, including exercise, education, thermal and electrical agents, and orthopedic manual therapy (OMT). Orthopedic manual therapy comprises a heterogeneous set of interventions such as thrust (TM) and non-thrust manipulation (NTM), soft-tissue manipulation, and neurodynamic movements. The use of OMT in clinical practice is supported across a wide body of literature [1–5] and has been endorsed by clinical practice guidelines (CPGs) [6–11]. Despite this evidential support, it is well known that results differ markedly across patient populations that are treated with OMT [12–18]. Additionally, specific application of OMT techniques may not result in superior outcomes when compared to randomly selected or nonspecific application of said techniques [19–23].

The aforementioned findings involving OMT are similar to those of other efficacious interventions. Clinical studies on chronic pain syndromes such as fibromyalgia and osteoarthritis have shown great inconsistency in therapeutic responses between individuals, even when efficacious interventions such as exercise or medications are used [24–26]. Variability in therapeutic response is partly explained by psychosocial factors and individual differences in how pain is modulated at the patient level27 and the failure to precisely characterize the pathophysiologic mechanisms that influence subgroups of patients [27]. This has prompted researchers to attempt the subgrouping of patients to better identify responders and non-responders to dedicated interventions: this is a concept known as phenotyping.

Phenotyping has been defined as ‘the ensemble of observable characteristics displayed by an organism’ [28]. Historically, phenotyping was associated with an individual’s interactions between their genotype and the environment. Only recently have researchers incorporated the use of clinical phenotyping, attempting to define groups of patients who may respond to specifically targeted treatments. Clinical phenotypes are based mainly on observable clinical and physiological measures. Clinical ‘pain’ phenotyping (whereby described as pain phenotyping) characterizes patients based on their pain experience and their response to dedicated interventions [28,29–31].

With respect to OMT, it has historically been assumed that the clinical effectiveness of the approach was associated with the technique selected (specialized training) and the skill of the clinician who applies the technique. We theorize that the differences in responses seen in clinical practice are reflective of variations in pain phenotypes and less reflective of techniques used or provider skill. While evidence suggests that higher-level training promotes better outcomes, we propose that this is related to improved ability to identify individuals who will have a favorable response to care rather than improved skill of OMT application [32]. The purpose of this paper is to reinforce the assumption that the likelihood of achieving a positive outcome with an OMT approach depends on selecting a patient with an amenable pain phenotype. To support our theory, we will present work supporting: (1) Different OMT techniques provide similar and consistent mechanisms in preclinical research; (2) current non-OMT pain phenotyping approaches support the assumption that patient-related factors are associated with a given pain response; (3) pain profiling in clinical trials has failed to consistently identify individuals with favorable outcomes to OMT; and (4) with the appropriate research, clinicians have the capacity to identify potential pain phenotype in their clinical assessment.

Different OMT techniques provide similar and consistent mechanisms in preclinical research

The term ‘mechanism’ reflects the steps or processes through which intervention (independent variable) actually unfolds and produces the change (outcome variable). The National Institutes of Health defines mechanistic studies as designed to understand a biological or behavioral process, the pathophysiology of a disease, or the mechanism of action of an intervention [33].

Neurophysiological mechanisms of all OMT forms have been well established as a significant driver of the analgesic response [34–39]. Neurophysiological mechanisms combine to create a peripheral, spinal, and supraspinal modulatory response which is thought to be the cornerstone for pain inhibition [34,40]. Peripheral pain inhibition in response to OMT has been demonstrated with a local increase in pain pressure threshold (PPT) without the same response remotely [41]. It is proposed that this peripheral mechanism is attributed to a modification in inflammatory markers [34,40,42]. All forms of OMT reduce inflammatory mediator expression leading to increased PPT and reduction in temporal summation [43–45].

Changes in PPT locally without remote changes are not consistent; therefore, when congruent changes occur locally and remotely, a central inhibitory effect is suggested [41,46]. This effect is proposed to be related to changes in pain-modulating peptides and neurotransmitters influenced by OMT. Serotonin and dopamine are altered with OMT [35] and act as pain-modulating neurotransmitters altering the affective component of pain [47]. The proposed sites of action for both dopamine and serotonin are widespread, including the dorsal horn of the spinal cord, periaqueductal gray, thalamus, basal ganglia, insular cortex, and cingulate cortex [48,49]. Oxytocin is another pain-modulating peptide affected by OMT [35], mitigating pain at the brain and spinal cord level [50]. At the spinal level, reviews on both humans and animals have established the effect OMT has on improved descending inhibition [35,51,52]. Studies on temporal summation, a measure of dorsal horn excitability, support direct inhibition with OMT [53–55].

Autonomic nervous system (ANS) response to OMT is assessed through measuring skin temperature, skin conduction, heart rate, and cortisol level changes [40]. A recent review concluded that OMT techniques affect the ANS with a combination of sympathetic and parasympathetic nervous system reactions [42]. The same review was not able to find evidence to support any differences in ANS response with differing manual techniques; however, this was not measured in most of the included studies [42].

Pre-clinical research involving TM and NTM produces consistent mechanistic responses across animal and human populations including transient neurophysiological (peripheral and central pain modulation), biomedical (immune and inflammatory systems and musculoskeletal stiffness), and psychological (contextual) mechanisms. These findings are supported by a number of systematic reviews involving preclinical research [35,41,43,51,52,56]. Although it is clear that the translation of these mechanisms to clinical outcomes is woefully understudied, it is unlikely that differences in mechanisms alone account for the great variability we see in clinical outcomes reported by patients.

Current pain phenotyping approaches support the assumption that patient-related factors are associated with a given pain response

Phenotyping studies use cluster and latent class analysis to identify characteristics of those with favorable analgesic response. These analyses identify patterns in multiple dependent variables and correlate them to related outcome variables to develop groups. Pain phenotyping trials have used both predictors of pain state (pain intensity, temporal summation, and conditioned pain modulation) and response to analgesic management as outcomes within these models.

Cross-sectional designs develop phenotypic groups based on characteristics demonstrated by those with optimal versus impaired pain states [57–62]. Grouping by conditioned pain modulation and temporal summation demonstrates the interaction between pain intensity, locations, and threshold contributing to phenotype, while psychological variables did not contribute [62]. Grouping patients with osteoarthritis based on the persistence of knee pain demonstrates the interaction between PPT and temporal summation contributing to phenotype, while similarly no interaction from psychological variables is demonstrated [59]. Grouping by self-reported functional limitations, pain intensity, gait speed, and health-care utilization demonstrates the interaction between pain intensity, locations, threshold, psychological distress, and number of comorbidities contributing to phenotype [61]. Grouping LBP based on pain intensity and variability demonstrates the interaction between fatigue, pain qualities, and presence of central sensitization contributing to phenotype [60]. Grouping patients with chronic pain based on pain characteristics (intensity, location, and duration) demonstrates the interaction between psychological strain, comorbid fibromyalgia diagnosis, social distress, and gender contributing to phenotype [57]. Grouping based on the presence of somatic widespread symptoms demonstrates the interaction between anxiety, depression, and emotional stability contributing to phenotype [58].

Phenotyping based on responsiveness to analgesic intervention has also demonstrated contributions from patient-specific characteristic [63–64]. Grouping patients with chronic pain receiving multimodal inpatient care based on reported measures of pain burden shows presence of depression, previous pharmacological and psychological intervention, and patient age all interacting to contribute to phenotype [63]. Individuals with high pain burden at baseline demonstrate improved pain outcomes, while those in other subgroups of pain burden show only improvement in depression and anxiety outcomes [63].

Higher baseline pain intensity is phenotypic for treatment effect size following lumbar spinal manipulative therapy (SMT) [65]. Patient expectations of recovery and comfort during treatment are phenotypic for response following thoracic SMT, while biomechanical parameters of SMT do not alter response [66]. Grouping based on local response to sustained noxious stimulation (pain adaptability) has identified two phenotypes both in healthy controls [64] and in patients with musculoskeletal pain67: those who are pain adaptable (demonstrate a reduction in pain over time) and those who are non-pain adaptable (demonstrate no reduction in pain over time). Latency to peak pain was the only variable that significantly correlated with pain adaptability. In these studies [64,67], conditioned pain modulation did not contribute to adaptability phenotype, suggesting another rationale for analgesia independent of central pain modulation.

The use of pain profiling in clinical trials has failed to consistently identify individuals with favorable outcomes to OMT

Clinical trials [68–71] and reviews [72–74] support OMT’s direct effect on reducing pain sensitivity locally and remotely. This is consistent regardless of the persistent or episodic nature of pain [75]. These findings are not specific to one technique with similar findings with manipulation, mobilization, and sham procedures [68,69,76]. Increased sensitivity at baseline correlates with larger increases in PPT; however, this association did not translate to between-group difference in clinical pain outcomes [71,72].

Early clinical studies attempted to associate changes in hypoalgesia with clinical response following TM [71,77,78]. Remote and local changes in PPT were not correlated with clinical SMT response in the short and medium term [78]. This corroborates previous work finding minimal correlation between changes in pain sensitivity and clinical outcomes [79]. Improved local hypoalgesia was identified following SMT in individuals with favorable outcomes regardless of the level of TM, while only SMT targeting the most painful segment was correlated with improved hypoalgesia irrespective of response [71]. These findings suggest a segmental reflexive modulation response promoting the hypoalgesia [71]. Current literature has failed to consistently identify factors associated with favorable OMT response. The overall lack of correlation between pain profile and clinical pain outcome suggests that nonspecific underlying factors such as a patient’s pain adaptability phenotype contribute to the analgesic response. Clinical research should investigate factors such as pain adaptability in an attempt to better identify those who will have a favorable response to OMT.

With the appropriate research, clinicians have the capacity to identify potential pain phenotypes in their clinical assessment

Preliminary observational work suggests that the early positive response is indicative of better long-term outcomes [12,80,81]. In these studies, the authors have associated clinical characteristics identified early in the examination with improved outcomes at discharge and up to 6 months. Specifically, a 30% reduction of pain from baseline was identified with repeated posterior to anterior mobilizations, and this response led to a 2.5–5.0 increased odds of an improved outcome versus a lack of this finding. In each of the studies, OMT was a focus of the treatment approach, causing the authors to assume a link between the initial 30% and the ultimate clinical outcomes with an OMT treatment, as other influencing variables were not controlled. While these findings may appear promising, the studies failed to distinguish whether or not the early clinical findings are specific to OMT; at this point, the findings are purely prognostic. We hypothesize that the clinical characteristic identified early in the examination is related to the patient’s endogenous pain-modulating capacity. In other words, the positive outcome with an OMT approach depended on the individual patient with an amenable pain phenotype, a term known as pain adaptive behavior64,65

To confirm the association of the aforementioned response with pain adaptive behavior, one would need to complete a concurrent validity study that includes both a clinical assessment and a laboratory-based exploration of one’s endogenous pain modulation. One can investigate endogenous pain modulation using the cold pressor test, which explores an individual’s ability to respond to sustained or repeated noxious stimuli locally by either decreasing sensitivity to the stimuli (adaptable) or increasing sensitivity (non-adaptable) [64,67]. If the adaptability and 30% change with the posterior–anterior mobilization are concurrent and strongly associated, it provides initial evidence of a proxy clinical measure for identifying pain adaptability. However, to explore whether this phenomenon is unique to OMT, one needs to go a step further and perform a responder analysis.

Typically, responder analyses are performed concurrently with a parallel, randomized controlled trial, in which patients are randomized to receive the treatment of interest or a comparator. A responder’s analysis involves identifying the number of responders in the treatment of interest group and comparing them to the number of responders who are in the comparison group [82]. This design assumes a counterfactual comparator. Historically, researchers have identified percent or point change (or some other clinically important designations) as a milestone to identify whether someone was a responder or not. Unfortunately, this method is fraught with error.

The problem with a comparison of responders in parallel trials is that the counterfactual (the supposed comparison person in a parallel, randomized controlled trial) does not truly exist [83]. A crossover trial, a design in which the same patient receives both treatments (a true counterfactual), is the only true way of determining treatment superiority. In this type of trial (Figure 1), patients are randomized to either manual therapy or exercise as the first intervention, followed by a wash-out phase to remove the effects of that approach, before assessing the second intervention. An a priori determination of how strong of a difference between interventions is needed to determine who truly responded to one intervention and not another.

Figure 1. Cross-over trial needed for responder’s analyses.

A cross-over trial is extremely difficult to do, and it is especially challenging with musculoskeletal conditions. The greatest change in patient-reported outcomes occurs very early in the care of the patient. Patients assigned to either manual therapy or exercise first will notably change in comparison to those assigned to the interventions second, and it is questionable whether an adequate interventional wash-out can occur. Cross-over trials require careful mathematical adjustments to the data to truly distinguish responders; otherwise, the order effect of the care received is likely the strongest effect that will be identified in the calculations.

Conclusion

We proposed that the likelihood of achieving a positive outcome with an OMT approach depends on the patient having an amenable pain phenotype. We provided evidence that mechanisms associated with an OMT technique are similar across pre-clinical animal and human studies. We identified studies that have characterized pain phenotypes based on treatment response, pain outcomes, and patient characteristics. These findings emphasize the value of patient-specific factors in clinical outcomes. We promote rather than attributing OMT effectiveness to the technique applied clinicians attribute the response to the patient’s pain phenotype and therefore modifications to the plan of care should be based on this concept. We proposed two research study designs needed to assess the value of a clinical examination approach toward identifying a pain phenotype construct and then identifying whether OMT leads to unique responses in that specific pain phenotype. Our suggested designs should help control against the well-known moderating and mediating variables of the patient and patient–clinician interactions associated with OMT [11,12]. Further research is needed to verify our suggestions, but we argue that these are the most important questions that require answering for OMT research.

Disclosure statement

Additional information

Funding

The authors reported that there is no funding associated with the work featured in this article.