Stratified medicine is changing our concept of disease

Understanding the different molecular pathways behind a disease will help doctors to manage it.

Stratified medicine seeks to understand the mechanism behind diseases and put patients into groups that will respond to a certain treatment.

If three patients go to their doctor with the common symptoms of rheumatoid arthritis — pain, swelling and stiffness of the joints — all three are likely be put on the antifolate methotrexate. But, statistically, this drug will only be completely effective for one of them, and the other two will eventually have to be switched to more costly biologic therapies.

By the time a patient’s medication has been changed, some of the damage to the joints wrought by the arthritis may be irreversible, says John Isaacs, a clinical rheumatologist at Newcastle University. And we currently don’t have a way to predict how patients will respond to treatment.

There is a growing belief among researchers that part of the reason why patients diagnosed with the same disease respond differently to treatment is because they don’t actually have the same disease. Much as travellers may arrive at a destination using different modes of transport, it is now thought that patients may present with a similar collection of symptoms that are caused by different underlying disease pathways.

So in the future a patient may be identified as having a certain type of rheumatoid arthritis and so will respond to a certain drug

“At the moment, rheumatoid arthritis is thought of as one disease but it is probably hundreds of diseases,” says Isaacs, who is leading the RA-MAP project at Newcastle University, which is attempting to identify the various immune pathways involved in rheumatoid arthritis.

Isaacs explains that the idea is to understand more about the individual patient’s disease — its severity and the treatments that are most likely to be effective. “So in the future a patient may be identified as having a certain type of rheumatoid arthritis (dubbed endotypes) and so will respond to a certain drug.”

Accurate diagnosis and treatment of disease is central to medical practice. But even when the diagnosis is right, there can be vast differences in the way common illnesses affect different people. Part of the answer lies in understanding a person’s genes, but this is not the whole picture; the environment can be an important driving force too. Researchers are using data-rich ‘omics’ technologies to search for biological signs of the underlying pathology, and this might help predict which patients will respond to a particular treatment.


Different strokes

One of the earliest examples of separating a condition according to its underlying pathology occurred back in the 1930s when the distinction was first made between type 1 diabetes (insulin deficiency) and type 2 diabetes (insulin resistance)[1]
. It took some time to fully characterise the molecular pathways, which happened around the turn of the millennium[2]

Much progress has been made since the 1930s. It is now already commonplace to tailor cancer treatments to the characteristics of a patient’s tumour. “Personalised healthcare has revolutionised treatment for cancer. It’s hard to imagine treating breast cancer without first identifying whether a patient is HER2 positive to decide whether they should receive Herceptin,” says Clodagh Beckham, country medical lead at Swiss pharmaceutical company Roche.

It would be nice to see other diseases catching up with cancer, she adds. For the pharmaceutical industry, there has been a shift away from one size fits all drugs. A quarter of Roche’s phase 2 and phase 3 drugs pipeline consists of immunology, inflammatory and ophthalmology drugs, and the vast majority of these are being evaluated with a companion diagnostic, or a biomarker is being sought. “After cancer, the next wave of stratified medicines will be in immunology and inflammation. Current treatments are just not doing enough — patients still have severe disease or flare ups,” says Beckham.

In non-cancer diseases, a move to separate patients based on molecular pathways has only really taken off in the last four or five years, supported by cheaper genome sequencing and statistical data analysis to interrogate linked clinical and multi-omic data.

In an autoimmune disease such as rheumatoid arthritis, there can be many different causes for the common symptoms of the illness and many of the affected pathways are shared with different autoimmune diseases. Stratified medicine seeks to understand the mechanism behind these causes and put patients into groups that will respond to a certain treatment.

“We have a shortage of symptoms with which to diagnose patients,” says Jonathan Pearce, stratified medicine programme manager at the UK Medical Research Council (MRC), which will invest £60m in this area by the end of 2014, including the RA-MAP project involving Newcastle University and several other UK institutions. The ultimate goal, says Pearce, is to understand the disease mechanism behind the different response and use this to develop new therapeutic strategies.

Crude condition

Like rheumatoid arthritis, asthma is probably an umbrella term for patients with different types of inflammation leading to similar symptoms. Approximately 10% of patients have severe disease that doesn’t respond to standard therapy. In a report by the Association of the British Pharmaceutical Industry (ABPI), it is estimated that, on average, any one class of asthma drug is only effective for 40% of patients (see graph). “There are multiple types of inflammation that can cause asthma and this is why different drugs work,” says Tim Harrison, clinical associate professor at the Faculty of Medicine and Health Sciences, Nottingham University. With the arrival of new monoclonal antibodies for asthma, understanding the different mechanisms that cause the symptoms of asthma has never been so important because these drugs are expensive and specific. “You only want to use them in the right patients. This way the treatment is more effective as well as more cost-effective,” says Harrison.

In an effort to better control severe asthma, UK pharmaceutical company GlaxoSmithKline developed a monoclonal antibody, mepolizumab, that targets interleukin 5 — a protein expressed by eosinophils, immune cells that play an important role in inflammation in some types of asthma. “When it was first tested, they gave it to a whole group of patients with severe asthma and it had no effect,” says Harrison. When a patient has inflammation caused by eosinophils, lots of these cells are seen in the sputum, he explains. It used to be assumed that all asthma patients would have eosinophilic inflammation, but after sputum testing was brought in, it became clear that some did and some didn’t. However, when given to patients with ongoing eosinophilic inflammation, it worked very well[3]
. So a better understanding of the inflammatory nature of asthma has led to new treatments and the recognition that patient selection is important, he says. Harrison believes there are probably three or four types of inflammation that lead to asthma. “Eosinophilic is the classic allergic asthma. But for patients with non-eosinophilic inflammation it is usually neutrophilic, which may be a consequence of chronic infection of the innate immune system.”

But it’s not just cutting edge drugs that elicit different responses in different patients. Salbutamol, the most commonly used treatment for asthma, is associated with different levels of effectiveness in different populations of patients. Esteban Burchard, principal investigator at the asthma origins laboratory at University of California, San Francisco, says that subtypes of asthma have been recognised for a few years now but that most treatments have only been tested in a narrow group of patients.

“In the United States, asthma is most prevalent and has the highest mortality in black and Puerto Rican patients,” says Burchard. In 1999, his research group found that a particular genetic variation, which increases asthma severity in white patients, was 40% more prevalent in black patients[4]
. Eight years later, his team discovered that salbutamol is less effective at expanding the airways of black and Puerto Rican patients with moderate and severe asthma

. He believes that poor response to treatment may be a contributing factor to mortality and morbidity in these patients. “Social factors also play a role but we believe genetics could play an even bigger role,” he says.

Burchard explains that 96% of genome-wide association studies (GWAS) have been conducted in people of European origin in genetic terms. So advances in understanding the role that genes play in disease through GWAS do not necessarily extend to other ethnicities.

If we have complete understanding of the disease pathways then this allows development of effective drug targets

His team is now recruiting 10,000 asthma patients from various ethnic groups to participate in a study of salbutamol response. The researchers will perform genetic analysis as well as taking detailed clinical measurements.

“Scientifically, it is important to understand how drugs work in all populations, not just in Europeans. If we have complete understanding of the disease pathways then this allows development of effective drug targets,” says Burchard (see ‘Clinical trials — the right people’).

Gut feeling

Researchers are also trying to separate out the different pathways that lead to inflammatory bowel disease (IBD). Dermot McGovern, director of translational medicine at the Inflammatory Bowel and Immunobiology Research Institute at Cedars Sinai Medical Centre, Los Angeles, says IBD is a “perfect storm” for personalised medicine.

We need to be able to predict how severe the disease will be in different patients so that we can target them with different therapies

Ulcerative colitis and Crohn’s disease fall under the banner of IBD but even within these subtypes there is a lot of clinical heterogeneity and a number of monoclonal antibodies for these conditions are being launched, he says. The course of inflammatory diseases can vary greatly between individuals but there has been real progress in understanding them. More than 160 different areas of the genome have been identified that contribute to a patient’s susceptibility to IBD and researchers are now doing “fine mapping” of these genetic regions in order to define the actual causative variants. “We need to be able to predict how severe the disease will be in different patients so that we can target them with different therapies,” he says.

McGovern, like Burchard, says that an essential part of his research is understanding ethnic differences. “Historically, IBD has been a disease that primarily affects Europeans and particularly the Ashkenazy Jewish population. But now we’re seeing an ‘epidemic’ in East Asia, and also a rising prevalence of the disease in other populations including African-Americans,” he explains.

In East Asian populations, for example, a gene called TNF alpha super family 15 (TNFSF15) is strongly associated with Crohn’s disease

but while the TNFSF15 signal is seen in Europeans, it is not as strong, he says. However, a monoclonal antibody is now being developed against this protein. In the future we may be able to test for the genetic variants of this gene, says McGovern, and target the therapy to patients whose genotype suggests that this pathway is important in driving their disease.

Populations with a high burden of the disease, like Ashkenazi Jews, are useful to study, he adds. “In the wider disease population, the smaller, incremental effects of these genes may be missed.” You can learn a lot from small, high-risk populations, he says, because the gene variants have such a pronounced effect, meaning they’re easier to identify, which can help unpick the biology of the disease.

An extreme example of this is early onset IBD, which develops in children a few years after birth. “We think this might be a slightly different disease,” says McGovern. Defects in the anti-inflammatory cytokine, interleukin 10 (IL10) pathway, have been implicated in early onset IBD

. Affected children have genetic variants in this pathway that have large effects on the function of IL10, explains McGovern. They are resistant to almost all existing therapies — the only effective treatment is a bone marrow transplant, he says. This knowledge can also be applied to the adult population with IBD.

“These data can help us understand how more subtle genetic variants in adults can increase the risk of disease and also potentially tell us about which therapeutic approaches we should be adopting in these patients,” he says.

McGovern’s research is trying to identify subgroups that will benefit from certain medicines. But he says that because IBD is genetically complex “we are beginning to accept that it’s not usually going to be one genetic variant that helps us achieve this goal”.

Genetic crossover

To add to the complexity of IBD, it also shares susceptibility genes with other inflammatory conditions. “One thing that will happen is a move away from a diagnosis of IBD, Crohn’s or ulcerative colitis, and towards a more molecular diagnosis that may spill over into other conditions,” says McGovern. IBD has the most overlap with ankylosing spondylitis — a chronic inflammatory condition affecting the spine and other areas of the body — but there is also overlap with type 1 diabetes and psoriasis. However, sharing of genes is not necessarily straightforward. One variant can increase the risk of IBD but decrease the risk of other conditions whereas other variants may increase the risk of both, says McGovern. Consequently, using the right drug becomes complicated. “Drugs like ustekinumab, which is for psoriasis, may also help IBD patients. But other psoriasis drugs, like secukinumab, may make it worse,” he explains.

In IBD, it is important to understand how the patient’s immune system is affected by their microbiome — the complete ecological profile of a person’s microbes. McGovern believes the increase in IBD in East Asia is probably linked to the adoption of a more Western lifestyle.

IBD occurs in genetically susceptible individuals who have an inappropriate immune response to the microbial flora in their gut. “The connection with the microbiome is really complex and researchers are trying to work it out,” he says. By looking at all the metabolites, or small molecules, in the gut, which include those produced by microbes, we can try to identify how these affect the immune system and how that may tie back to the individual’s genetic makeup, he adds. “For example, if you had a microbiome that led to a difference in the amount of short chain fatty acids, some individuals may have a particular genetic variant that makes them very susceptible to that change and therefore to developing inflammation.” Answering these complicated questions is causing researchers to look at more than just genetics. McGovern says that patients will probably be grouped based on a number of combined factors, including serology (antibodies in the serum), microbiome data and genetics, along with clinical and demographic characteristics.

The data

The opportunities to stratify patients are increasing as genomic sequencing gets cheaper and the ability to bring -omics data together in a computational way reveals opportunities. “We are on the cusp of being able to integrate all this information together,” says Watkins at the MRC. “Combining different types of data will require sophisticated algorithms but the opportunities to achieve this are increasing.”

However, Pearce warns that algorithms are likely to give a “probabilistic” prediction, which could be difficult for patients to accept. “When there are a limited number of options to offer patients, communicating why you’re not giving them a drug is going to be a challenge. Most of the time there isn’t going to be a definitive yes or no answer to whether or not they’ll respond to a drug.”

Isaacs’ consortium, studying rheumatoid arthritis, is also looking for a biological signature that goes beyond the patient’s genes. “We’re only analysing 300 patients, which is a relatively small sample size for genetics alone. So we’re using the -omics technologies: transcriptomics (mRNA), metabolomics (metabolites), proteomics (proteins), as well as genomics, to better understand the abnormalities in the immune system that lead to the disease that we recognise as rheumatoid arthritis.”

The main challenge, he says, is making sense of the millions of data points, which need to be combined and then analysed — no one has done something quite like this before. “The aim is that in 20–25 years you will be able to run a set of blood tests and after analysis an answer will pop out saying this is the type of rheumatoid arthritis this patient has and this is the most appropriate treatment.”

As Pearce puts it: “Right now disease is defined anatomically, but we are moving to a future in which disease will become defined by mechanism.”

Panel: Clinical trials — the right people

Understanding disease pathways requires analysis of a broad spectrum of patients. Currently, clinical trials of drugs do not necessarily include a representative population. Tim Harrison, clinical associate professor at the Faculty of Medicine and Health Sciences, Nottingham University, says clinical trials don’t really reflect the patient population that we see every day. For example, for patients to be accepted on clinical trials for asthma, they often have to have a 12–15% reversibility in lung function and this excludes 60% of asthma patients, he says. “No trials include smokers and then we use these drugs in smokers and wonder why they don’t work.”

This concern extends to a lack of ethnic diversity within trials. Esteban Burchard, principal investigator at the asthma origins laboratory at University of California, San Francisco, says: “[Ethnic minorities] are patient populations that are not widely represented in clinical trials and so when these drugs are used in these patients they may not work or they may have unexpected adverse reactions. This is especially obvious when looking at the way patients of different ethnic backgrounds respond to different asthma therapies.”

Regulatory authorities do not necessarily mandate the inclusion of data from ethnic minorities or other disease populations in clinical trials. When it comes to inflammatory bowel disease (IBD), McGovern says that now the disease is becoming more prevalent in the Asian populations, there is an increasing need to represent these groups in all kinds of research. “However it is increasingly difficult to recruit enough patients for IBD clinical trials in North America and Western Europe, so adding regulatory requirements in these regions could limit patient recruitment and hence delay drugs reaching the market, he says. Nonetheless, he says there seems to be a concerted move to make research more inclusive. “A lot of advances have been made, and while it used to be almost all European ancestry-based datasets, my group is generating data in Puerto Ricans, African Americans and Korean populations and this is also true of many groups worldwide,” says McGovern.

Jonathan Pearce, stratified medicine programme manager at the UK’s Medical Research Council, argues that a balanced approach needs to be taken when designing trials. He says that the use of “overly restrictive” inclusion criteria in a clinical trial can result in disease subtypes being excluded. “However, the adoption of overly broad inclusion criteria can result in an efficacy signal in one of the subtypes being lost due to the increased noise,” he adds.

The answer could be in collecting real-world data. Harrison says that this is already happening. “Rather than clinical trials changing, we’re seeing a wave of new types of trial.” There is a big move to run later phase ‘real-life’ pragmatic studies in the general population, which are complementary to traditional trials.

And once patient populations and disease mechanisms have been identified, this in turn can make clinical trials more efficient with the knowledge of which patient group the drug is likely to work in.


1.Himsworth HP. Diabetes mellitus: its differentiation into insulin-sensitive and insulin-insensitive types. Lancet 1936;1:127–130.

2.Cavaghan MK, Ehrmann DA & Polonsky KS. Interactions between insulin resistance and insulin secretion in the development of glucose intolerance. J Clin Invest. 2000;106:329–333.

3. Nair P, Pizzichini M, Kjarsgaard M et al. Mepolizumab for prednisone-dependent asthma with sputum eosinophilia. N Engl J Med. 2009;360(10):985–993.

4.Burchard EG, Silverman EK, Rosenwasser LJ et al. Association between a sequence variant in the IL-4 gene promoter and FEV(1) in asthma. Am J Respir Crit Care Med. 1999;160(3):919–922.

5.Naqvi M, Thyne S, Choudhry S et al. Ethnic-specific differences in bronchodilator responsiveness among African Americans, Puerto Ricans, and Mexicans with asthma. J Asthma. 2007;44(8):639–648.

6.Yamazaki K, McGovern D, Ragoussis J et al. Single nucleotide polymorphisms in TNFSF15 confer susceptibility to Crohn’s disease. Human Molecular Genetics. 2005;14(22):3499–3506.

7.Glocker E-O, Kotlarz D, Boztug K et al. Inflammatory bowel disease and mutations affecting the Interleukin-10 receptor. N Engl J Med 2009;361:2033–2045.


Last updated
Clinical Pharmacist, CP, December 2014, Vol 6, No 10;6(10):DOI:10.1211/PJ.2014.20067355

You may also be interested in