Richard Smith: Stratified, personalised, or precision medicine

Richard SmithDoctors know that many of the patients they treat with drugs will not benefit. Many patients know that too, which may be why some don’t take their drugs. The simple concept behind stratified medicine is that we become smarter at identifying the patients who will benefit. What has long been an aspiration is slowly becoming a reality, but transforming health systems, developing regulatory processes, and finding business models to promote stratified medicine is proving tough. The Academy of Medical Sciences thus held a workshop last week to try to overcome the barriers to improve patient treatment and allow both British academia and UK PLC to flourish.

I was invited to contribute my ignorance to the workshop, and I must confess that when first asked I didn’t know what stratified medicine was. I quickly learnt that it could be seen as personalised medicine, which I had heard of, rebranded—because personalised medicine has got a bad name through being oversold. It’s perhaps fairer to say that stratified medicine (identifying groups of patients who will benefit from treatments) is a step towards personalised medicine, when treatments will be tailored to individuals, probably as a result of sequencing their entire genome. The US Institute of Medicine has used the term precision medicine and as always it would be good to agree on what we call it.

The end of diseases as we know them

Precision medicine is a good term because current medicine is imprecise. For centuries medicine classified diseases by what could be seen, felt, and smelt, and then about a century ago we began to move to defining diseases anatomically, physiologically, and biochemically. This is where most of medicine is now, but even in 1944 the cardiologist Thomas Lewis said that “Diagnosis is a system of more or less guessing in which the outcome is a name.” We can now do better using molecular biology and genetics, and relatively soon we will move to individualised computer prediction. Familiar diseases like breast cancer, diabetes, Alzheimer’s disease, and rheumatoid arthritis will become many diseases.

Because diagnosis is so imprecise it follows that treatment will be haphazard, and we heard at the meeting of an account from Finland of 509 women with ovarian cancer treated with 124 different drugs in combinations that had little rationale. There is big variation in how individuals respond to drugs and yet that variation is not usually recorded. Drugs are approved on their average performance. The future, said one speaker, is not therapeutics and diagnostics but “theragnostics,” combining the two: how patients respond to treatments will be one of the main ways of classifying their disease.

Examples of stratified medicine working

There are now many examples of where a diagnostic test can limit the number of people who needed to be treated. One of the first examples is trastuzumab, better known as herceptin, a monoclonal antibody that interferes with the HER2/neu receptor and is used to treat breast cancer. If women do not have overactive HER2 receptors then the drug is not useful, and diagnostic tests approved by the US Food and Drug Administration (FDA) allow the identification of women who will not benefit from treatment.

Oncotype Dx tests  for 21 genes in tissue from breast cancers and allows the identification of a subset of women with breast cancer who do not need chemotherapy. This means that the women can avoid the side effects of chemotherapy and the health system save the costs of treatment. The FDA approved Oncotype Dx on the basis of a retrospective examination of datasets of women with breast cancer. The National Institute of Health and Clinical Excellence (NICE) is currently considering Oncotype Dx, and the company that produced the test argues that at a price of $3500 the cost per quality adjusted life year (QALY) is around ₤9000, whereas NICE puts it at between  ₤30 000 and ₤40 000, a cost at which it may not be approved. There is as yet no decision.

Several new drugs are available for treating advanced malignant melanoma, which has been untreatable, and vemurafenib (which is marketed as Zelboraf) inhibits the B-Raf enzyme and works only in patients who have the BRAF V600 mutation. Interestingly for me who worked on apoptosis, programmed cell death, in 1973 the drug works by promoting apoptosis (I’ve been waiting 40 years for a real world use). The drug and the companion diagnostic test were approved by the FDA in four months, the fastest approval ever.

Most of the current examples of stratified medicine are in cancer and involve expensive drugs. The tests do not identify people who will have an enhanced response but rather those who will not respond. At the moment stratified medicine is more what will be than what is. Most current examples are a combination of a single drug and test, but rapidly we will move to multiple tests and multiple treatments. Indeed, it’s likely to be less than a decade before the whole genome of individuals will be sequenced.

How can we prepare for such a world? That was the main topic of conversation at the Academy meeting, and the barriers are formidable. “Economists are criticised for caring only whether something works in theory, but I can’t get this to work even in theory,” said one economist who has been working on stratified medicine for years, both in industry and academia.

Barriers to stratified medicine: economic

The economics are an important part of the problem. At the moment a drug company might be paid, say, a ₤100 a year per patient for treating 100 patients, giving them an income of ₤10 000. The number needed to treat might be 10, meaning that only 10 of the 100 patients benefit from taking the drug. Now suppose somebody develops a test that allows identification of the 10 who benefit. The health system might be delighted to pay for the test (and the test might even be developed within its own labs, so no company needs paying) and for the drug for the 10 people who will benefit. The income for the drug company would drop to ₤1000, providing a powerful disincentive to any company to develop ways of identifying the patients who will benefit. Indeed, currently drug companies are like mustard makers in that they are making most of their money from people who do not benefit from their drugs—just as mustard makers are making the most of their money from what diners leave on their plates.

In the above case the drug company might argue that more value has been created for people and the health system in that 90 people will be avoiding the side effects that all drugs have and the 10 who are taking the treatment can enjoy what economists call “the value of knowing.” It also seems likely that the 10 who know that they will benefit from the treatment will have better adherence than the 100. So the company might argue that it should be paid more than the ₤10 000 it was paid for treating the 100, especially if it also manufacturers the diagnostic test. Many drug companies are now developing diagnostic arms (some have them already), and if they can patent both a new drug for an incurable condition together with a diagnostic test then they put themselves in a very powerful position.

Sometimes the diagnostic test and the drug will be developed together by one company, but the diagnostic test may be developed by another company (or even a hospital or academic lab) and after the drug is on the market. The only way to cope with all circumstances is to have value based pricing where the price of a drug can go up as well as down. This is hard stuff for payers of healthcare.

Barriers to stratified medicine: regulation

Regulators are also struggling with how to approve combinations of drugs and diagnostic tests, and it seems increasingly that companies will be seeking licences for combinations rather than single drugs. The FDA regulates treatments and diagnostic tests, but Europe has separate systems. How much evidence is needed and who will pay for it? Diagnostic companies have a very different business model from drug companies. Usually the tests are cheap and the product lifetime short, so the companies don’t have either the financial resources or the skills to conduct the studies that will be needed by regulators. Drug companies so far have paid for the evidence needed by regulators, but there will be no incentive for them if it means fewer people taking their drugs and no increased payment.

And what sort of evidence should be needed? Ideally there should be randomised trials showing reduced mortality and morbidity with one trial to show the benefit of the test and another the benefit of the drug, but this double randomisation is “an unaffordable luxury,” said one health economist. Randomised trials for all combinations will clearly be unrealistic when we arrive in the world of multiple tests and multiple treatments.

Barriers to stratified medicine: upgrading the NHS

But what about the health system itself, perhaps the NHS? The NHS is clearly not ready for such a world when most healthcare workers have never heard of stratified medicine and are trained for a world of diseases that will be rendered redundant and ways of treating that will be subdivided. There is clearly a huge job to be done in upskilling the workforce, and it’s a matter not only of skills but also of attitudes and culture, which are harder to change.

In addition, a world in which diseases fragment and there are multiple tests and treatments will demand a capacity to handle vast amounts of data. These databases will need both clinical and research utility to justify investment. Electronic records linked to multiple and quality assured databases will be essential, but there is also a big need for clinical informaticists who can ensure that the data is available, high quality, and well managed. Such people are different from bioinformaticists and simply don’t exist in Britain at the moment.

The public will also have to be introduced to this brave new world, and the simple idea that they will be given treatments that will be more effective for them (and that they won’t have to take ineffective treatments) is attractive. It will be important for people to be willing to have their data (including perhaps their genomes) included in large databases that can be used for research. The data will, of course, be anonymised, but people will have to understand that as multiple databases are inked together the chance increases that people might be identified. Various surveys show that most people are willing for their data to be used, and a debate is underway about including an opt out rather than opt in clause in the NHS constitution for making data available to researchers, including drug companies.


So stratified medicine creates great scientific and therapeutic excitement and beckons a very different world, but there are formidable barriers to reaching that new world. One of the co-chairs of the committee that produced the American report on precision medicine put it thus: “Developing this new network and the associated classification system will require a long-term perspective and parallels the challenges of building Europe’s great cathedrals—one generation will start building them, but they will ultimately be completed by another, with plans changing over time.”
I must say that I left the meeting with my head spinning and wondering about the future of medicine where most people in the world don’t now get simple drugs that could benefit them, people are living longer, welfare states are becoming unaffordable, and we are exhausting the planet—but I’ll address all that in another blog,

You can learn more about the meeting by looking at RS’s Tweets at #stratified medicine

Competing interest: RS attended the meeting, contributed (not very much) to a paper prepared in advance for the meeting, and had his accommodation and food paid for by the sponsors of the meeting, which included various companies and governmental bodies.

Richard Smith was the editor of the BMJ until 2004 and is director of the United Health Group’s chronic disease initiative.

  • Huw Llewelyn

    This is indeed very exciting. However, my understanding of medical progress
    is that it progressively makes medicine more stratified, personalised and
    precise. For example, at one time we
    diagnosed ‘dropsy’ and prescribed digitalis (NNT about 100 perhaps). We then ‘stratified’ ‘dropsy’ by working out in a more precise way that it was those with ‘dropsy’ AND a rapid irregularly irregular
    pulse who respond best (NNT 1 to 2 perhaps). We called it ‘atrial fibrillation’! We now have a choice of treatments e.g. beta-blockers,
    cardio-version, warfarin, asprin, etc that will suit some patients more than
    others depending on other factors, thus ‘personalising’ the medicine too.

    The Appendix of the Oxford Handbook of Clinical Diagnosis explains
    to students how all this can be done in a logical and systematic way by using
    symptoms, signs and tests (which includes genetic information). It also explains why using two standard
    deviations to select patients for treatment is inappropriate, using the albumin
    excretion rate as an example to show how cut-off points can be placed in a
    pragmatic way. So, we must also see these developments in the context of the
    history of medicine.

  • kidmugsy

    Yet I routinely see people urging me to attain a particular BMI without mentioning my age, my sex, or my race. Stratified medicine would seem to be the opposite of what the Medicalbullying Tendency seeks to achieve.