The idea of science based physiotherapy (adapted from the popular science based medicine) is a response to the notion that evidence based physiotherapy places too great an emphasis on comparative clinical studies at the expense of basic science. This is to say that evidence based physiotherapy undervalues, or even ignores aspects such as mechanisms, mechanistic reasoning and biological plausibility. This call for a greater emphasis on improving our mechanistic reasoning and our understanding of how and why we achieve the outcomes that we do is important. Sound mechanistic reasoning can improve the manner in which we research and deliver interventions while helping rule out more far-fetched ideas prior to dedicating limited scientific resources to them. This is because it encourages us to ask the question “are we confident that this treatment aligns with current knowledge of biology, physiology and physics?” As with all things, of course, it is important to recognize the limitations of our mechanistic reasoning as means of justifying our treatments.
Jeremy Howick describes mechanistic reasoning as taking a peek “inside the black box” to see what actually happens when an intervention is delivered. More explicitly, he writes “mechanistic reasoning involves an inference from mechanisms to claims that an intervention produces a patient-relevant outcome.” One of the more well known and unfortunately disastrous examples mechanistic reasoning is illustrated by Howick in his book, Philosophy of Evidence Based Medicine:
Myocardial infarction often damages the muscle and electrical system in the heart, leaving it susceptible to arrhythmias. A common type of arrhythmia, ventricular extra beats (VEBs), occurs when the left ventricle contracts before it has had time to fill completely. The heart then fails to pump sufficient blood. Without treatment, lung, brain, and kidney damage ensues. Worse, VEBs can also degenerate into ventricular fibrillation, or complete electrical chaos. Sudden death soon follows ventricular fibrillation in the absence of electric shock. Large-scale epidemiological studies suggested that between 25 and 50% of sudden cardiac deaths were associated with arrhythmias. Based on this understanding of the underlying mechanisms, several drugs were developed and found to be successful for regulating VEBs. The drugs became widely prescribed in the belief that they would reduce cardiac deaths ... However, a randomized trial suggested that the drugs increased mortality, and had killed more people every year than died in action during the whole of the Vietnam War.
The example above illustrates the difficulty of mechanistic reasoning in physical therapy and medicine — the remarkable complexity of human biology and physiology. There are countless other examples of mechanistic reasoning that at the time appeared intuitive and physiologically sound based upon available knowledge, but actually ended up being completely wrong and and some tragic cases, harmful. This overwhelming complexity coupled with the human tendency to oversimplify can lead to significant errors in decision making when reasoning is based on what Howick describes as an empty or partial understanding of the mechanisms involved.
Over the last decade, evidence has been mounting that the surgical lavage and debridement of painful arthritic knees actually provides no meaningful benefit to patients when compared to a sham surgery. This is despite a seemingly intuitive rationale that lavage and debridement “cleanses the knee of painful debris and inflammatory enzymes [and relieves symptoms through the] removal of flaps of articular cartilage, torn meniscal fragments, hypertrophied synovium, and loose debris.” However, this line of reasoning fails as it is based on a partial understanding of both knee osteoarthritis and the mechanisms of pain. Such reasoning is myopic and focused on linear, single system thinking which subsequently has created what is estimated to be a 3 billion dollar annual cost to US healthcare, much of which could be argued as wasteful.
Homeopathy serves as an example of empty mechanistic reasoning due to the fact that potentization, a central tenet of producing homeopathic treatment, has no evidential basis and is far beyond the realm of biological plausibility. Homeopathy’s lack of grounding in any current understanding of physics and human biology is shown by David Gorski when he writes:
Thanks to some basic laws of physics and chemistry and a little thing known as Avogadro’s number, any homeopathic dilution greater than 12C (twelve serial 100-fold dilutions) is incredibly unlikely to contain even a single molecule of starting compound. That unlikeliness reaches truly astonishing levels as we reach the common homeopathic dilution of 30C, which is the equivalent of a 1060-fold dilution. Given that that little thing known as Avogadro’s number, which describes how many molecules of a compound are in a mole, is only approximately 6 x 1023, a 30C dilution is on the order of 1036– to 1037-fold higher than Avogadro’s number. Even assuming that a homeopath started with a mole of remedy before diluting (unlikely, given the high molecular weight of most of the organic compounds that can serve as homeopathic remedies), the odds that a single molecule could remain behind after the serial dilution and succussion process is infinitesimal.
In physical therapy, there has been a recent swing in the proposed mechanism(s) for many of our interventions, from a strictly mechanical and structural basis towards a predominantly neurophysiological explanation. This has been a mostly welcomed change towards being less wrong, as many of our previous structural explanations for treatments have had difficulty holding up under scientific scrutiny. However, it is important to be cautious with our enthusiasm for the neurophysiological explanations of particular treatments given our likely partial understanding of the processes involved. As Jeremy Howick writes: “Our knowledge of pathology and physiology is rarely sufficiently complete to infer precisely how an intervention will affect mortality or morbidity.” An observed change in a certain physiological measure after administering a treatment is not necessarily the specific mechanism of action in which the treatment works. This goes back to the complexity of the system — the change measured certainly could be how the treatment works, but it could also simply be an innocuous correlatory result of the treatment, or it could be a complete red herring. Acknowledging the fact that our current interest in things such as cortical smudging, novel inputs, biochemical markers, mechanoreceptors and neuroimaging could be just as erroneous as our previously cherished structural rationale is to adopt an essential aspect of professional humility and uncertainty.
Manual therapy and other similar interventions have been popularly described as novel, non-threatening neurophysiological input altering a patient’s perception which very well might be true but in explaining them in such a way, we propose a mechanism so vague that it becomes meaningless. Nearly everything can be described as a novel neurophysiological input altering a patient’s perception — whether we are talking about rubbing cabbage, applying magnets, manual therapy, exercise or even a slap in the face. This, as Paul Mitalski says, leaves us with a definition that is “over ambitious and under precise.” Caution should also be exercised when describing interventions as working through sharpening cortical smudges and other more specific claims, absent sufficient evidence that such a process is occurring and clinically relevant to the treatment at hand. Otherwise, we might be making the same mistakes that were made with extravagant structural claims, now still quite wrong, just in a different way. To be useful, our mechanistic reasoning needs to demonstrate to the best of our current ability, a sound understanding of the events that occur from “mechanisms to claims that an intervention produces a patient-relevant outcome.” Joel Bialosky and colleagues created a model for manual therapy to improve our understanding of the mechanisms involved and guide future studies in hopes of avoiding reliance on vague, incomplete or erroneous mechanistic reasoning. Their work serves as an excellent example of the need for empirical testing of “hypotheses related to different biomechanical and neurophysiological effects specific to types of manual therapy” in order to improve our mechanistic reasoning.
As it stands, mechanistic reasoning in physical therapy probably best serves as a means of ruling out treatments, or treatment explanations, that extend well beyond the realm of possibility to conserve finite scientific resources and aid in the application of interventions to individuals. Mechanistic reasoning can be use to instantly preclude homeopathy from needing to undergo randomized controlled trials and other testing to elucidate its benefit. To continue to perform trials studying the efficacy or effectiveness of homeopathy is an unfortunate waste of time and resources due to the biologically bankrupt nature of its purported mechanism of action and formulation. Though, the idea of ruling out based off of mechanistic reasoning has pitfalls as well — Howick provides examples of instances where partial understanding of mechanisms led to the delays in introduction of many useful interventions such as the initial ridicule faced by those who suggest H. Pylori was a predominant cause of peptic ulcers limiting the adoption of antibiotics as a treatment. Therefore, we must approach clinical problems from both a mechanistic and a clinical research lens to achieve a clearer understanding.
The basic science that informs our mechanistic reasoning must co-exist with real world comparative clinical trial data. This is because the physiological measures and surrogate outcomes studied in basic science and early exploratory work may or may not actually influence outcomes that patients care about. The knowledge that manual therapy may cause a decrease in central nervous system temporal summation or alter serum levels of cytokines is somewhat meaningless if it does not translate to an improvement in someone’s pain or disability. The use of surrogate outcomes to sell a treatment’s usefulness is seen often in the pharmaceutical and CAM industries — There are countless examples of novel medicines introduced to market with exciting claims of improving a particular surrogate, such as lowering A1c in diabetes, but failing to actually influence real world outcomes such as mortality or risk of cardiovascular event. When pairing the elegant information obtained from basic science with comparative clinical trials measuring meaningful endpoints, we are able to obtain a more complete understanding of why a treatment works, how it works, who it will work for and how meaningful the effect actually is.
The takeaway here is that we must be humble in the claims and inferences made from mechanistic reasoning. Humans are remarkably complicated and the unknown of the systems that human beings are made up of probably outweigh the known. It is seductive and alluring to flesh out the black box that exists between intervention and observed outcome, but to do so with certainty is incredibly difficult due to the limitations in our understanding of human systems and the probabilistic nature of mechanistic reasoning. Overconfidence in mechanistic reasoning at best misleads us into providing ineffective treatments and at worst can cause significant harm. Humility should be exercised when discussing how and why our treatments might work (or not work). There should be an increased comfort with uncertainty and a move away from over-confident explanations with tenuous levels of support or rationale.
Photo courtesy of flickr user nerovivo