Inseparable bedfellows: serendipity and research
On the train-ride back from the Joint Annual Medical Research Society/Academy of Medical Sciences/Royal College of Physicians Clinician Scientists in Training Day, I was feeling inspired about basic science research, which does not happen often enough to epidemiologists like me. Every year they invite young researchers to present their findings and invite eminent speakers to encourage the next generation to stick with research. It is rare to go to a meeting where scientists from all disciplines from genetics and haematology to cardiology and epidemiology all present their work, but this meeting works.
I heard about some genuine “bench to bedside” research, from MRI scans to gauge neural function in intensive care patients to novel treatments for liver cirrhosis and metastatic breast cancer. The overall Young Investigator Prize winner was a very deserving Tamir Rashid from Cambridge. His research group has firstly produced stem cells from skin cells, which raises the possibility of not needing donor cells or embryonic cells in future. They have then developed new technology to correct the error in the genetic code in a form of liver cirrhosis (alpha-1 antitrypsin deficiency) using stem cells with the correct code. Finally, they have shown that these cells restored liver function in humans and the therapy is about to be tested in a clinical trial. This is a brilliant example of “bench to bedside” research starting with a hypothesis, testing in animal models and then developing a strategy of treating disease in humans, later to be tested in a trial. But how does research like this happen?
In plenary speeches, Professor Sir John Tooke and Professor Andrew Hattersley repeatedly stressed four important aspects of successful clinical research: good mentors, collaboration, serendipity and patient involvement. Professor Charles Warlow, the eminent stroke researcher, has written that “Good luck should be exploited, and very often barriers can be not just overcome but put to good use as well”.
Turning to Pubmed or to Google, we find countless examples of serendipity in scientific discovery and research. Luck has been important in the development of drugs, whether for cardiovascular diseases, malaria or depression. One of the most fascinating examples of good fortune and medical discovery in my own area of cardiology is the accidental injection of the right coronary artery by Mason Soanes in 1958 which led to the development of coronary angiography, and most of the advances in modern cardiology in the last half-century.
The social scientists refer to healthcare organisations and research organisations as complex adaptive systems(CASs) and as such, serendipity will always play a major role. A recent study looked at researchers' perspectives on “how high impact publications are developed and why they are consistently produced by a small group of researchers”. Interestingly, the researchers consistently found five factors: (1) rules of thumb; (2) these rules of thumb were reinforced by positive feedback from peers and mentors; (3) good communication skills allowing researchers to provide feedback to their peers, thus closing a positive feedback loop; (4) curiosity, open-mindedness and motivation; and (5) randomness and serendipity. The study found, unsurprisingly, that some researchers were better than others at capitalising this randomness. So the message for aspiring researchers, from the available evidence, is to be in the right place at the right time,stick with it if you get lucky, and don't forget to talk to your research team!
EBM at the bedside-bicuspid aortic valves and familial screening
The original proponents of EBM have always argued for “evidence at the bedside” so that we can make the best decisions for patients nearest to the point “where the rubber hits the road”. How often do we clinicians actually look up the evidence in real time during or soon after a consultation to change the management or the advice we give to a patient?
I saw a lady in her 40s in our cardiology clinic this week. She has been followed up every 1-2 years in clinic for bicuspid aortic valve (BAV). Basically, the aortic valve is at the outflow of the left ventricle (the major pump of the heart) and usually has three cusps which open and close to ensure flow of blood in the right direction through and out of the heart. In bicuspid valves, people are born with only two cusps and over their lifetime, they are more prone to developing narrowing of the valve (“aortic stenosis”), with a significant probability of needing aortic valve replacement during their lifetime. The idea of screening and surveillance is that any narrowing or malfunction of the aortic valve can be picked up early, and the person can be referred for surgery more quickly and effectively than if their disease had progressed.
BAV is the most common abnormality of the heart valves, occurring in 1- 2% of the general population and is twice as common in males as in females. Reassuringly, a recent cohort study of patients with BAV found that they have similar survival rates to the normal population. However, “given that serious complications will develop in over a third of patients with BAV, the bicuspid valve may be responsible for more deaths and morbidity than the combined effects of all the other congenital heart defects”. The potential problems are narrowing or leaking of the aortic valve, infective endocarditis and enlargement or “dilatation” of the aorta. In other words, BAV is common, has serious complications and there is a treatment which improves survival (aortic valve replacement). Therefore, BAV is a condition which meets Wilson’s criteria for screening.
I was asked by the lady if her children were at risk of BAV and whether they should be screened. I did not know the exact answer so I looked online with the patient. There is a 30% risk of aortic dilatation or BAV in first degree relatives (parents, children or siblings) of people with BAV. A more recent study showed that 20% of first degree relatives of people with BAV may have undetected BAV themselves. It turns out there are no NICE guidelines or formal UK/European guidelines for whether we should be screening relatives or how we should be doing it.
Interestingly, across the pond, the Americans have guidelines for familial screening and the literature seems to suggest it. Therefore adult children of patients with BAV should have an echocardiogram to check that they do not have a BAV which would mean that they should also be followed up. Valvular heart disease is a bigger health issue than we imagine.
There are four take home messages for me. First, EBM can be done at the bedside-it is meant to be the most practical of clinical sciences. Second, there is no harm as a clinician in saying “I don’t know” and looking it up. Third, sometimes it is the obvious clinical questions which are still unanswered or debatable. Finally, practice can be changed.
Weighing up benefit and harm -net clinical benefit and subgroup analysis
The Hippocratic oath originally included the harm and good that doctors and their prescribed treatments can cause. The biggest challenge in today’s clinical practice is not much different. With increasing numbers of trials of different drugs in different patient groups with different comparison groups, how are patients and doctors ever going to see the wood from the trees? How do we make judgments about which drug to use in which situation?
NICE was set up in 1999 in order to help in these difficult matters. Broadly speaking, it looks at current trial evidence and uses the metrics of “cost-effectiveness” to decide whether to fund drugs and treatments in the NHS. It uses “quality-adjusted life years” (the ‘QALY') to measure effectiveness and then calculates the cost per QALY gained for a given drug. A drug must be effective in treating disease but the cost of the benefit must be below a certain threshold, usually £20000-30000 per QALY gained
One problem is that in trials, we tend to focus on benefits and not harms. Another problem is that the performance of drugs in different patients, even for simple characteristics like age and sex and poorly defined in many trials. Even more importantly, trials often do not report their outcomes based on the disease risk of the patients involved. Therefore we end up “painting all patients with one brush”. This has obvious problems. Cost effectiveness analysis is only as good as the trials which are studied and if those trials do not report outcomes (good and bad) properly, then analysis is difficult.
Atrial fibrillation (AF) is a heart rhythm problem which causes increased risk of stroke. Warfarin has been established as a safe treatment for over 50 years and reduces risk of stroke. However, it does lead to increased risk of bleeding, including intracerebral bleeds. Therefore, a way of quantifying the overall benefit of warfarin is to directly weigh up the risk of stroke and the risk of intracerebral bleeds as a “net clinical benefit”, as proposed by Singer and his colleagues in 2009. They reported that “Expected net clinical benefit of warfarin therapy is highest among patients with the highest untreated risk for stroke, which includes the oldest age category.” In other words, we should use the drug in the patients with the highest chance of benefit from the drug, or the highest chance of the adverse outcome (intracerebral bleeds).
Currently 3 new drugs (dabigatran, apixaban and rivaroxaban) have been evaluated in trials as alternatives to warfarin in the setting of AF. Each of these trials looks at different patients and uses different comparisons. In a recent analysis, we used data from the Danish National Patient registry to work out the net clinical benefit of these drugs at different levels of risk of stroke (potential benefit) and bleeding (potential harm) compared with warfarin. We also calculated the number of patients needed to treat and harm for each drug at each level of risk. Although, this is a modelling exercise, this type of analysis is needed in order to look at all the drugs side by side, using the best evidence we currently have. This idea of “net clinical benefit” could also be used in other disease areas in order to quantify to both health professionals and patients how good or bad a treatment is.