The Laws of Life: Theory and Evidence

1. Understanding mechanisms

Mechanisms and their relevance to clinical practice can only be properly interpreted if their scientific underpinnings are understood. And the road to understanding many mechanistic principles in medicine is well approached through an appreciation of the foundational chemical principle known as the Law of Mass Action and the various biological relationships, which I shall call “the laws of life”, that proceed from it.

In this series of short articles, and using practical examples as illustrations, I shall outline these laws of life, as follows:

1. Guldberg and Waage’s Law of Mass Action
2. The Henderson–Hasselbalch equation
3. The Michaelis–Menten equation
4. The Langmuir adsorption isotherm
5. Ligand-receptor binding
6. Concentration-effect curves and dose-response curves
7. Transport equations
8. Signals and messengers
9. Feedback and control mechanisms
10. Translating signals and messages into clinical outcomes

The mechanisms described by these processes relate to physiological, biochemical, and pharmacological processes.

As an introduction to the series, I shall start with a discussion of the role of mechanisms in relation to clinical evidence.

2. Evidence and Mechanism-Based Reasoning

The “Levels of Evidence” recommended by the Oxford Centre for Evidence-Based Medicine (CEBM) include “mechanism-based reasoning”, ranked as the lowest level of evidence for four of the six categories covered (diagnosis, treatment benefits, treatment harms, and screening). The downgrading of mechanism-based reasoning in this way is understandable, since there are cases in which mechanisms have been used to generate hypotheses that were subsequently proved to be wrong.

However, as with the inclusion in early forms of the evidence hierarchy of “expert opinion” as the lowest form of evidence, one could argue that this is a category error. Expert opinion in clinical practice is not a form of evidence, but is ideally a set of beliefs based on a synthesis of all the best relevant evidence combined with experiential judgment about what is likely to be best for an individual patient. Expert opinion should not be confused with clinical expertise, which is the ability of experts to apply their clinical skills, clinical experience, and evidence-based knowledge to the management of practical problems, modified in the light of the views and preferences of others, such as the patient and the patient’s carers.

Similarly, mechanism-based reasoning does not constitute evidence in the way that the other categories included in the levels of evidence (such as systematic reviews, randomized studies, non-randomized studies, and case series) do. This is reflected, for example, by the fact that one can adduce evidence for mechanisms themselves using the levels of evidence, which one cannot do for the other levels. This is illustrated in the figure below, which diagrammatically shows that the term “evidence of mechanisms” is not the same as “mechanisms as evidence” (often referred to as “mechanistic evidence”). Here I use the former to mean the evidence that demonstrates and elucidates mechanisms and discard the latter, preferring the term “mechanisms for interpretation”.

Figure. Different forms of clinical evidence (top orange box) can lead to conclusions of clinical relevance (ochre box); both clinical and non-clinical evidence (orange boxes) can provide evidence of mechanisms (orange arrows); mechanisms (right-hand blue box) can be used to interpret clinical conclusions (mechanisms for interpretation; purple arrow); clinical expertise is defined in the text.

Parenthetically, we can note that if mechanism-based reasoning were a form of evidence, there might be cases in which it could be used as strong evidence, in the way that anecdotal observations, usually regarded as poor evidence, can sometimes afford strong evidence in determining both beneficial  and harmful  effects of therapeutic procedures. Categorizing case reports as [always] being of poor evidential quality does not solve this problem. The GRADE system, by allowing certain levels of evidence to be upgraded in certain circumstances does offer some relief, but does not completely solve the problem. “Between-the-eyes” adverse reactions, for example, are of high evidential value but are still relegated to the bottom of the hierarchy, even in GRADE.

3. The true role of mechanisms

If mechanisms do not constitute evidence, what do they constitute? The answer is that they constitute information that can be used either to generate hypotheses or to help in interpreting evidence that arises from other sources. The higher the quality of the evidence underpinning the mechanism, the more useful mechanism-based reasoning becomes. Such evidence may include clinical or non-clinical evidence and the latter may include in vitro evidence, ex vivo or in vivo (i.e. animal) evidence, in silico evidence (e.g. computer simulations), and in cerebro evidence (i.e. thought experiments).

Consider two examples.

1. In vitro evidence that a drug inhibits a cytochrome P450 isoenzyme, such as CYP3A4, suggests that it will cause interactions in vivo with drugs that are metabolized by CYP3A4. This is not evidence that such an interaction will occur, but it suggests a hypothesis that needs to be tested in vivo, when evidence in either direction may be obtained. Conversely, if such an interaction is detected in vivo, subsequent in vitro evidence of a specific mechanism can add to its interpretation. In either case the evidence is not the mechanism but the enzyme inhibition that suggests the mechanism.
2. That a drug causes a reduction in total serum cholesterol concentration is evidence of a mechanism that might be associated with a reduction in the long-term risk of heart attacks and strokes, but a randomized controlled trial of the effect of the drug on those outcomes would be necessary to provide evidence of the hypothesized efficacy. In this case one would not do the RCT without the prior hypothesis based on mechanism. A strong mechanism, as in this case, provides a high prior probability, while a weak mechanism, as in the case of homoeopathy for example, provides a low prior.

Mechanism-based reasoning, like expert opinion, should therefore be placed outside the scheme of levels of evidence (see figure). The fifth level in the levels of evidence could be replaced by an item such as “Non-evidential information incorrectly treated as if it were evidence [of a similar kind to RCTs etc]”. This would include mechanisms and expert opinion as well as, for example, formal consensus, reports of expert committees, and basic principles. The levels of evidence could be applied to the experiments on which mechanisms themselves are based (i.e. evidence of mechanisms, as defined above). If a helpful mnemonic is required, this category could be called “Faulty evidential interpretation of generally non-evidential data” (FEIGNED) or “Non-evidential information that hasn’t an evidential role” (NEITHER).