Effectiveness of manual therapies: the UK evidence report

EBH is about doing the right things for the right people at the right time [5]. It does so by promoting the examination of best available clinical research evidence as the preferred process of decision making where higher quality evidence is available [6]. This reduces the emphasis on unsystematic clinical experience and pathophysiological rationale alone while increasing the likelihood of improving clinical outcomes [7]. The fact that randomized clinical trial (RCT) derived evidence of potentially effective interventions in population studies may not be translated in a straight forward manner to the management of individual cases is widely recognized [8‐10]. However, RCTs comprise the body of information best able to meet existing standards for claims of benefit from care delivery. The evidence provided by RCTs constitutes the first line of recommended action for patients and contributes, along with informed patient preference, in guiding care [11]. Practice, as opposed to claims, is inherently interpretative within the context of patient values and ethical defensibility of recommendations [8, 12]. Indeed, the need to communicate research evidence, or its absence, to patients for truly informed decision-making has become an important area of health care research and clinical practice [13, 14].

While some may argue that EBH is more science than art [7], the skill required of clinicians to integrate research evidence, clinical observations, and patient circumstances and preferences is indeed artful [6]. It requires creative, yet informed improvisation and expertise to balance the different types of information and evidence, with each of the pieces playing a greater or lesser role depending on the individual patient and situation [15].

It has become generally accepted that providing evidence-based healthcare will result in better patient outcomes than non-evidence-based healthcare [7]. The debate of whether or not clinicians should embrace an evidence-based approach has become muted. Put simply by one author: "...anyone in medicine today who does not believe in it (EBH) is in the wrong business [7]." Many of the criticisms of EBH were rooted in confusion over what should be done when good evidence is available versus when evidence is weak or nonexistent. From this, misunderstandings and misperceptions arose, including concerns that EBH ignores patient values and preferences and promotes a cookbook approach [16]. When appropriately applied, EBH seeks to empower clinicians so they can develop fact-based independent views regarding healthcare claims and controversies. Importantly, it acknowledges the limitations of using scientific evidence alone to make decisions and emphasizes the importance of patients' values and preferences in clinical decision making [6].

The question is no longer "should" we embrace EBH but "how"? With EBH comes the need for new skills including: efficient literature search strategies and the application of formal rules of evidence in evaluating the clinical literature [6]. It is important to discern the role of the health care provider as an advisor who empowers informed patient decisions. This requires a healthy respect for which scientific literature to use and how to use it. "Cherry-picking" only those studies which support one's views or relying on study designs not appropriate for the question being asked does not promote doing the right thing for the right people at the right time.

Perhaps most critical is the clinician's willingness to change the way they practice when high quality scientific evidence becomes available. It requires flexibility born of intellectual honesty that recognizes one's current clinical practices may not really be in the best interests of the patient. In some cases this will require the abandonment of treatment and diagnostic approaches once believed to be helpful. In other cases it will require the acceptance and training in new methods. The ever-evolving scientific knowledge base demands that clinicians be accepting of the possibility that what is "right" today might not be "right" tomorrow. EBH requires that clinicians' actions are influenced by the evidence [17]. Importantly a willingness to change must accompany the ability to keep up to date with the constant barrage of emerging scientific evidence.

The purpose of this report is to provide a brief and succinct summary of the scientific evidence regarding the effectiveness of manual treatment as a therapeutic option for the management of a variety of musculoskeletal and non-musculoskeletal conditions based on the volume and quality of the evidence. Guidance in translating this evidence to application within clinical practice settings is presented.

The available evidence usually includes consistent results from well-designed, well conducted studies in representative populations which assess the effects on health outcomes.

The evidence is based on at least one higher-quality randomized trial (low risk of bias) with sufficient statistical power, two or more higher-quality (low risk of bias) randomized trials with some inconsistency; at least two consistent, lower-quality randomized trials (moderate risk of bias). As more information becomes available, the magnitude or direction of the observed effect could change, and this change may be large enough to alter the conclusion.

For the purpose of this report a determination was made whether the inconclusive evidence appears favorable or non-favorable or if a direction could even be established (unclear evidence).

Additionally, brief evidence statements are made regarding other non-pharmacological, non-invasive physical treatments (for example exercise) and patient educational interventions, shown to be effective and which can be incorporated into evidence-based therapeutic management or co-management strategies in chiropractic practices. These statements are based on conclusions of the most recent and most updated (within last five to ten years) systematic reviews of randomized clinical trials and widely accepted evidence-based clinical guidelines (primarily from the UK and US if available) identified through our search strategy.

Translating evidence requires the communication of salient take-home messages in context of the user's applications [3]. There are two message applications for information derived from this work. First, the criteria for sufficiency of evidence differ depending on the context of the considered actions [8, 19]. Sufficient evidence to proffer claims of effectiveness is defined within the socio-political context [20] of ethics and regulation. Separate is the second application of evidence to inform decision making for individual patients. Where there is strength of evidence and the risk of bias is small, the preferred choices require little clinical judgment. Alternatively, when evidence is uncertain and/or there is higher risk of bias, then greater emphasis is placed on the patient as an active participant [11]. This requires the clinician to effectively communicate research evidence to patients while assisting their informed decision-making [19].

In summary, the information derived within this report are directed to two applications 1) the determination of supportable public claims of treatment effectiveness for chiropractic care within the context of social values; and 2) the use of evidence information as a basis for individualized health care recommendations using the hierarchy of evidence (Figure 1).

There are two important questions underlying the medical and media debate surrounding the scope of chiropractic care and claims regarding its effectiveness particularly for non-musculoskeletal conditions: 1) should health professionals be permitted to use generally safe but as yet unproven methods? 2) What claims, if any, can and should be made with respect to the potential value of unproven treatments?

In response to the second question, no claims of efficacy/effectiveness should be made for which there isn't sufficient evidence. Unsubstantiated claims can be dangerous to patient health [289]. We maintain the best evidence for efficacy/effectiveness that meets society's standards comes from well-designed RCTs. While other study designs and clinical observations do offer insight into the plausibility and potential value of treatments, the concepts of plausibility and evidence of efficacy/effectiveness should not be confused when making claims.

Why is it that the results of RCTs often do not confirm the results observed in clinical practice? There are several reasons. One of the problems is that both the provider and the patient are likely to interpret any improvement as being solely a result of the intervention being provided. However this is seldom the case. First, the natural history of the disorder (for example. acute LBP) is expected to partially or completely resolve by itself regardless of treatment. Second, the phenomenon of regression to the mean often accounts for some of the observed improvement in the condition. Regression to the mean is a statistical phenomenon associated with the fact that patients often present to the clinic or in clinical trials at a time where they have relatively high scores on severity outcome measures. If measured repeatedly before the commencement of treatment the severity scores usually regress towards lower more normal average values [290].

Additionally, there is substantial evidence to show that the ritual of the patient practitioner interaction has a therapeutic effect in itself separate from any specific effects of the treatment applied. This phenomenon is termed contextual effects [1, 291]. The contextual or, as it is often called, non-specific effect of the therapeutic encounter can be quite different depending on the type of provider, the explanation or diagnosis given [292], the provider's enthusiasm, and the patient's expectations [293‐298]. Some researchers have suggested that relying on evidence from RCTs and systematic reviews of RCTs is not adequate to determine whether a treatment is effective or not. The main issue, they contend, is that the intervention when studied in RCTs is too highly protocolized and does not reflect what is going on in clinical practice [230]. They advocate a whole systems research approach that more accurately represents the entire clinical encounter. When using this perspective and systematically synthesizing the literature regarding chiropractic treatment of non-musculoskeletal conditions, also reviewed in this report, they conclude, for example that chiropractic is beneficial to patients with asthma and to children with infantile colic [230]. This conclusion is at odds with the evidence summaries found in this report. We submit that whole systems research approach in this instance is clouding the interpretation of the literature regarding effectiveness as it relates to making claims, and incorrectly giving the consumer the impression that chiropractic care shows effectiveness over and above the contextual effects as it relates to the two examples above.

In a placebo-controlled RCT the question is: does the treatment provided have a specific effect over and above the contextual or non-specific effects. The result of such a trial may show that there is no important difference between the active intervention and the sham intervention. However, the patients may exhibit clinically important changes from baseline in both groups and thus the outcome would be consistent with what clinicians observe in their practice. An example of this is the results of the pragmatic placebo controlled RCT on chiropractic co-management of chronic asthma in adults (care delivered by experienced chiropractors consistent with normal clinical practice), which showed that patients improved equally during both the active and the sham intervention phases of the trial [231].

It is essential to recognize what each piece of the EBH puzzle offers. Patient values and preferences do not provide sound evidence of a treatment's effectiveness and may be misleading. A patient can be satisfied with a treatment, but it still may not be effective. The clinician's observations, if well documented, can attest to patient improvement while under care and encourage perception of a treatment's clinical plausibility. However, the narrow focus of attention under non-systematic observations common to practice experience tends to obscure other factors influencing case outcome. Similarly, EBH can be flawed, not because it fails to be scientific, but because-like all sciences-it imports the biases of researchers and clinicians [299]. Well-performed clinical research however, does provide evidence for claims that a treatment is effective when the results are consistently applied to relevant patients. This is because of its reliance on methods for systematic observation and efforts to minimize bias.

Other authors' work has been used to argue that a range of study types should be included when evaluating a treatment's efficacy/effectiveness (case series, etc.) [230, 300]. We maintain the best evidence that rises to societal standards to support claims of efficacy/effectiveness comes from well-designed RCTs. This is largely due to the powerful effect of successful randomization and design factors intended to minimize bias (all which help ensure that the results are due to the intervention and not some other known or unknown factor). Other evidence may be useful to inform treatment options when conditions for individual patients are not consistent with the best evidence or when better evidence is unavailable [11]. Other types of research are more appropriate for answering related questions including, but not limited to, safety or mechanistic plausibility. This can lead to the refinement of interventions, inform the design of clinical trials, and aid in the interpretation of clinical observations. Similarly, clinical data from epidemiological studies, case reports, and case series can suggest that a treatment is clinically plausible. That is, clinical observations demonstrate that it is possible that an intervention is effective. However, a gain in plausibility, biological or clinical, does NOT constitute proof of a treatment's efficacy in human populations. Conversely lack of proof (as demonstrated through well performed randomized clinical trials) does not exclude plausibility [301, 302].

Research on systematic reviews have taught us that individual studies can often lead to a conclusion very different from that of a systematic analysis of all available studies [3]. Moreover, the scientific process is a systematic means of self-correcting investigations that classically begin with observations and hypotheses that support plausibility and/or mechanisms. Ideally, these precede and inform the conduct of RCTs under conditions most likely to yield clear results, often referred to as efficacy studies. Separately, studies that emulate general practice conditions may be used to develop an understanding of effectiveness. Historically, the modern investigation of manual treatment methods represents an aberration in this process. With the advent of social support and funding for research at the end of the 20^th Century, there was an underlying presumption that the long-term practice of these methods provided a sound clinical wisdom on which to ground RCTs, bypassing mechanistic studies. The early emphasis on clinical trials has illuminated the gaps in understanding of appropriate indications for treatment, dosage and duration of care, consistency of treatment application, and the appropriate outcome measures to monitor results [11]. In response, funding agencies in North America have renewed research emphasis on the potential mechanisms of effect [303]. Data from this work is expected to inform future clinical research questions, and subsequently lead to well-grounded studies that are likely to yield more complete evidence regarding appropriate and effective care.

Choosing an intervention should always be tempered by the risk of adverse events or harm. Adverse events associated with manual treatment can be classified into two categories: 1) benign, minor or non-serious and 2) serious. Generally those that are benign are transient, mild to moderate in intensity, have little effect on activities, and are short lasting. Most commonly, these involve pain or discomfort to the musculoskeletal system. Less commonly, nausea, dizziness or tiredness are reported. Serious adverse events are disabling, require hospitalization and may be life-threatening. The most documented and discussed serious adverse event associated with spinal manipulation (specifically to the cervical spine) is vertebrobasilar artery (VBA) stroke [304, 305]. Less commonly reported are serious adverse events associated with lumbar spine manipulation, including lumbar disc herniation and cauda equina syndrome [304].

Estimates of serious adverse events as a result of spinal manipulation have been uncertain and varied. Much of the available evidence has been relatively poor due to challenges in establishing accurate risk estimates for rare events. Such estimates are best derived from sound population based studies, preferably those that are prospective in nature [304, 306].

Estimates of VBA stroke subsequent to cervical spine manipulation range from one event in 200,000 treatments to one in several million [307, 308]. In a subsequent landmark population-based study, Cassidy et al [309] revisited the issue using case-control and case-crossover designs to evaluate over 100 million person-years of data. The authors confirmed that VBA stroke is a very rare event in general. They stated, "We found no evidence of excess risk of VBA stroke associated with chiropractic care compared to primary care." They further concluded, "The increased risk of VBA stroke associated with chiropractic and PCP (primary care physician) visits is likely due to patients with headache and neck pain from VBA dissection seeking care before their stroke." In regards to benign adverse reactions, cervical spine manipulation has been shown to be associated with an increased risk when compared to mobilization [55, 310, 311].

Appropriately, the risk-benefit of cervical spine manipulation has been debated [304, 305]. As anticipated, new research can change what is known about the benefit of manual treatment for neck pain. Currently, the evidence suggests that it has some benefit [55]. It has been suggested that the choice between mobilization and manipulation should be informed by patient preference [55].

Estimates of cervical or lumbar disc herniation are also uncertain, and are based on case studies and case series. It has been estimated that the risk of a serious adverse event, including lumbar disc herniation is approximately 1 per million patient visits [312]. Cauda equina syndrome is estimated to occur much less frequently, at 1 per several million visits [312‐314].

The true incidence of serious adverse events in children as a result of spinal manipulation remains unknown. A systematic review published in 2007 identified 14 cases of direct adverse events involving neurologic or musculoskeletal events, nine of which were considered serious (eg. subarachnoid hemorrhage, paraplegia, etc.) [315]. Another 20 cases of indirect adverse events were identified (delayed diagnosis, inappropriate application of spinal manipulation for serious medical conditions). The review authors note that case reports and case series are a type of "passive" surveillance, and as such don't provide information regarding incidence. Further, this type of reporting of adverse events is recognized to underestimate true risk [315‐317].

Importantly, the authors postulate that a possible reason for incorrect diagnosis (for example. delayed diagnosis, inappropriate treatment with spinal manipulation) is due to lack of sufficient pediatric training. They cite their own survey [318] which found that in a survey of 287 chiropractors and osteopaths, 78% reported one semester or less of formal pediatric education and 72% received no pediatric clinical training. We find this particularly noteworthy.

The conclusions in this report regarding the strength of evidence of presence or absence of effectiveness are predicated on the rules chosen for which there are no absolute standards. Different evidence grading systems and rules regarding impact of study quality may lead to different conclusions. However, we have applied a synthesis methodology consistent with the latest recommendations from authoritative organizations involved in setting standards for evidence synthesis. Although we used a comprehensive literature search strategy we may not have identified all relevant RCTs, guidelines, and technology reports. Conditions for which this report concludes the evidence currently shows manual treatment to be effective or even ineffective, sometimes rests on a single RCT with adequate statistical power and low risk of bias. Additional high quality RCTs on the same topics have a substantial likelihood of changing the conclusions. Including only English language reviews and trials may be considered another limitation of this report leading to language bias; however, the impact of excluding non-English trials from meta-analyses and systematic reviews is conflicting [319, 320], and the incidence of randomized trials published in non-English journals is declining [321]. Another potential limitation of this report is the lack of critical appraisal of the systematic reviews and clinical guidelines included in the report. Systematic reviews and clinical guidelines can differ widely in methodologic quality and risk of bias [322]. While critical appraisal of the included reviews and guidelines would be ideal, it was beyond the scope of the present report. When drawing conclusions about relative effectiveness of different forms of manual treatments it is acknowledged that it has usually not been possible to isolate or quantify the specific effects of the interventions from the non-specific (contextual) effect of patient-provider interaction [291]. It was beyond the scope of this report to assess the magnitude of the effectiveness of the different manual therapies relative to the therapies to which comparisons were made. However, if moderate or high quality evidence of effectiveness was established the therapy was interpreted as a viable treatment option, but not necessarily the most effective treatment available. We recognize that findings from studies using a nonrandomized design (for example. observational studies, cohort studies, prospective clinical series and case reports) can yield important preliminary evidence on potential mechanisms and plausibility of treatment effects. However, the primary purpose of this report is to summarize the results of studies designed to specifically address treatment efficacy and effectiveness from which claims of clinical utility, consistent with that literature, may be considered defensible. Therefore, the evidence base on the effects of care was restricted to RCTs.