23rd July 2018

Trying to work out what causes any disease is tricky, very tricky, although in some cases it has been relatively straightforward – at least in retrospect. Scurvy, for example, has a single cause. A lack of vitamin C. If you replace the vitamin C, all the signs and symptoms of scurvy will disappear.

Equally, tuberculosis, is caused by the single pathogen, or microorganism, the tuberculous bacillus. The discovery of the bacillus was made by Robert Koch in 1882 using his meticulous scientific technique, based on his famous postulates:

  • The microorganism must be found in abundance in all organisms suffering from the disease but should not be found in healthy organisms.
  • The microorganism must be isolated from a diseased organism and grown in pure culture.
  • The cultured microorganism should cause disease when introduced into a healthy organism.
  • The microorganism must be re-isolated from the inoculated, diseased experimental host and identified as being identical to the original specific causative agent.

Koch didn’t just stumble across a bacteria in someone with TB and announce to the world that this was the cause of TB. He knew that if you look in any sample from diseased lungs, you will find hundreds of different bugs kicking about. Which of them is the true cause?

To find out, you need to isolate one, find a culture where it can multiply, then stick it in another animal to see if it develops the same disease. You take that microorganism back out of the newly diseased animal and check it is the same bacteria that you isolated in the first place. Then, and only then, can you claim you found the causal agent.

Good stuff, sounds complicated. In truth, that was simple.

Things become far more difficult when, for example, you cannot find a single causal agent. Or you find that you have found a likely agent, but many people exposed to it do not get the disease. Or, you find that people who have not been exposed to your proposed causal agent can also get the same disease.

Smoking, for example. You have a hypothesis that smoking causes lung cancer, but most people who smoke do not get lung cancer. Equally, many people who have never smoked can suffer from lung cancer. Given this, you could argue that it is not actually smoking that causes lung cancer, but something else. An argument used for decades by the tobacco industry to establish that smoking was perfectly healthy.

Recognising these difficulties, in 1965 the English statistician Sir Austin Bradford Hill proposed a set of nine criteria, known as ‘Bradford Hills cannons of causation’. They were designed to provide a model for epidemiologic evidence of a causal relationship between a presumed cause and an observed effect. It was Hill and Richard Doll who demonstrated the connection between cigarette smoking and lung cancer. The list of the criteria, or cannons, is as follows:

Strength: (effect size): A small association does not mean that there is not a causal effect, though the larger the association, the more likely that it is causal.

Consistency: (reproducibility): Consistent findings observed by different persons in different places with different samples strengthens the likelihood of an effect.

Specificity: Causation is likely if there is a very specific population at a specific site and disease with no other likely explanation. The more specific an association between a factor and an effect is, the bigger the probability of a causal relationship.

Temporality: The effect has to occur after the cause (and if there is an expected delay between the cause and expected effect, then the effect must occur after that delay).

Biological gradient: Greater exposure should generally lead to greater incidence of the effect. However, in some cases, the mere presence of the factor can trigger the effect. In other cases, an inverse proportion is observed: greater exposure leads to lower incidence.

Plausibility: A plausible mechanism between cause and effect is helpful (but Hill noted that knowledge of the mechanism is limited by current knowledge).

Coherence: Coherence between epidemiological and laboratory findings increases the likelihood of an effect. However, Hill noted that “… lack of such [laboratory] evidence cannot nullify the epidemiological effect on associations”.

Experiment: “Occasionally it is possible to appeal to experimental evidence”.

Analogy: The effect of similar factors may be considered.

You may have noted that these cannons are not remotely black and white. There are many shades of grey here. Even so, I can confidently assure you that if you take any of the current risk factors for heart disease, they fail to meet some, many, or indeed any, of Bradford Hills cannons for causation.

For some time, I looked at Koch’s postulates, changing the word microorganism to pathogen. I reviewed the Cannons for causation and repeatedly tried to apply them to possible causes of cardiovascular disease, but I found that they are of little practical use. Things got very complicated very quickly and trying to pull all the necessary strands of thought together was well beyond my mental capacity.

I began to realise that when it comes to cardiovascular disease we do NOT have any single causal agent, or factor, or even a remotely coherent causal model. Something noted over twenty years ago.

‘Our poor understanding of the nature of coronary heart disease explains why we lack a clearly expressed paradigm to explain it. All diseases are explained on the basis of a paradigm, or model, which is an expression of present understanding even though it might be incomplete or wrong. Being able to develop a paradigm, to construct a model, implies a certain level of understanding; the absence of such a paradigm which would include most if not all known risk indicators, implies very little understanding.

In practice there is what can be regarded as a ‘flat paradigm’ for the development of coronary heart disease, in that it is thought to be due to the addition of a wide range of risk indicators. The flat paradigm of CHD means that it might appear to be due to genetic influences in one person, cigarette smoking in another, a faulty diet in another, a metabolic abnormality in another etc.. This contravenes traditional pathological teaching that a given disease has a specific cause, although a variety of factors might influence the natural history of the disease. In fact, the flat paradigm is simply a summation of observations and make no attempt to explain how the various factors might interact.’1

In that article, Grimes uses the term ‘flat paradigm’ which I rather like, but have never seen used before, or since. Others commonly describe CVD as being multifactorial, as though this helps in any way. ‘Yes, CVD is multifactorial.’ In reality, the use of this term is basically an admission of failure. ‘We don’t really know what causes CVD, but here is a list of things that we think might have something to do with it, in some people, but not alllook, stop asking difficult questions.’

This lack of any coherent model is reflected in the latest CVD risk calculator developed in the UK. It is called Qrisk3. There were two earlier models Qrisk1 and 2. You can bring up the calculator on-line and input all your ‘risk factors’. It will then work out your risk of having a cardiovascular ‘event’ in the next ten years – allegedly. It can be found here:https://qrisk.org/three/

It has twenty variable factors. These are

  • Age
  • Sex
  • Smoking
  • Diabetes
  • Total cholesterol/HDL ratio
  • Blood pressure
  • Variation in two blood pressure readings
  • BMI
  • Chronic kidney disease
  • Rheumatoid arthritis
  • Systemic Lupus Erythematosus (SLE)
  • History of migraines
  • Severe mental illness
  • On atypical antipsychotic medication
  • Using steroid tablets
  • Atrial fibrillation
  • Diagnosis of erectile dysfunction
  • Angina, or heart attack in first degree relative under the age of 60
  • Ethnicity
  • Postcode

There is an alternative calculator used in the US. It only uses ten factors. It can be found here http://www.cvriskcalculator.com/ As an aside, neither of them use LDL to calculate risk. Interesting? [Both calculators greatly overestimate risk].

Looking at the UK version, what does it all mean? Does this list imply understanding? No, it is just a mis-mash of the most common things that have found to increase your risk of CVD. Some of the items on the list can only be associations e.g.:

  • Age
  • Sex
  • Ethnicity
  • Postcode (Zipcode)
  • Angina, or heart attack in first degree relative under the age of 60

You may say that age clearly does cause CVD – it is certainly the most powerfully weighted factor on the list. I would counter that, if you have no other identified risk factors for CVD, why should getting older be a problem? What is the mechanism?

At least three items on the list are caused by CVD

  • Variation in two blood pressure readings
  • Erectile dysfunction
  • Chronic kidney disease

One of them sits completely alone

  • Atrial Fibrillation

As for the others:

  • Smoking
  • Diabetes
  • Blood pressure
  • Total cholesterol/HDL ratio
  • BMI
  • Rheumatoid arthritis
  • Systemic Lupus Erythematosus
  • History of migraines
  • Severe mental illness
  • On atypical antipsychotic medication
  • Using steroids tablets

Here we have ten causes? But can these extremely disparate things all cause the same disease, and in the same way. History of migraines, and smoking, for example – what links them. Or, severe mental illness and rheumatoid arthritis. Go on, try and fit them together, with the LDL hypothesis, and see if you can end up with a coherent model.

Therein lay the challenge that I set myself many years ago. Of course, I could easily add many other items to Qrisk3. Antiphospholipid syndrome, sickle cell disease, Kawasaki’s, air pollution, magnesium deficiency, Avastin, proton pump inhibitors and on and on.

The reality is that, when you analyse Qrisk3, it is immediately apparent that there is no single necessary and sufficient causal agent to be found here. One alternative would be to suggest that there are several hundred different varieties of CVD, all with their own specific cause, and all leading to the same pathophysiology [a term used to describe the disordered physiological processes associated with disease or injury.]

To put this another way, the classical causal models were never going to work for CVD. Koch’s postulates, Bradford Hills cannons for causation represent a paradigm that is not suitable for understanidng CVD. For starters it is impossible to establish how they can all fit together as independent factors.

Even if you restrict yourself to the twenty different variables on Qrisk3, the possible combinations between them is twenty factorial. Which is twenty times nineteen, times eighteen, times seventeen etc. That is 2,432,902,008,176,640,000 possible interactions. Two-point four sextillion. Go on, design a clinical trial to explore that.

What I came to realise, eventually, is that you cannot understand CVD by studyng hundreds and hundreds of different risk factors that could be causal, could be associations, could be coincidence. The only possibly way to understand this disease, was to stop looking for causes and start looking at the process. This is something that I have said many times before, but it bears almost endless repetition, for it is key to everything.

If you cannot explain why, and more importantly how such things as: postcode, rheumatoid arthritis, smoking, steroids and history of migraines can lead to an increased risk of CVD you are just making lists and explaining nothing. So, having got that off my chest, again, I shall return to process in the next instalment.

1: Grimes D, Hindle E, Dyer T: ‘Respiratory infection and coronary heart disease: progression of a paradigm.’ Q J Med 2000: 93:375-383

Source link