The earliest biomarkers, such a body temperature or blood pressure, were single measurements that reflected multiple physiological processes.  Today, though, our reductionist approach to biology has turned up the resolution of our lens: we can measure levels of individual proteins, metabolites and nucleic acid species, opening the biomarker floodgates.

But this increased resolution has not necessarily translated into increased power to predict.  The principal use of biomarkers after all is to use things that are easy to measure to predict more complex biological phenomena.   Unfortunately, the levels of most individual molecular species are, on their own, a poor proxy for physiological processes that involve dozens or even hundreds of component pathways.

The solution is to combine individual markers into more powerful signatures.   Biomarkers like body temperature allow physiology to perform the integration step.  But for individual molecular biomarkers that job falls to the scientist.

Unsurprisingly, the success of such efforts is patchy – simply because there are an infinite number of ways to combine individual molecular biomarkers into composite scores.  How do you choose between linear and non-linear combinations, magnitude of coefficients and even at the simplest level which biomarkers to include in the composite score in the first place?

The first port of call is usually empiricism.   Some form of prior knowledge is used to select an optimal combination.  For example, I may believe that a couple of biomarkers are more likely to contribute than others and so I may give them a stronger weighting in my composite score.   But with an infinite array of possible combinations it is hard to believe that this approach is going to come anywhere close to the optimum combination.

Unless you have a predictive dataset, however, this kind of ‘stab in the dark’ combination is the best you can do.  Just don’t be surprised if the resulting composite score is worse than any of the individual biomarkers that compose it.

With a dataset that combines measurements of each individual biomarker and the outcome being modeled, more sophisticated integration strategies become possible.  The most obvious is to test each individual marker in turn for its association with the outcome and then combine those markers that show a statistically significant association.  Perhaps you might even increase the weighting of the ones that are most strongly associated.

But how powerful are these ad hoc marker composites?

From a theoretical perspective, one might imagine the answer is not very powerful at all.  While common sense suggests that each time you add another marker with some new information in it the predictive power of the composite should improve, unfortunately this simple view is too, well, simple.   Each new marker added into a composite score contributes new information (signal) but also further random variation (noise).  To make a positive contribution, the additional signal has to be worth more than the additional noise.

Even when the data is available, asking whether each marker is significantly associated with outcome to be predicted is therefore only looking at one half of the equation: the signal.  It does little to quantify the noise.  Worse still, it doesn’t address whether the signal is “new” information.  Too often, the individual markers used to construct a composite are correlated with each other, so the value of each new marker is progressively reduced.

In sharp contrast, the random noise from different markers is rarely, if ever, correlated.  So each added marker contributes a full shot of noise, but a heavily diluted dose of signal.   Making biomarker composites more powerful than the best single marker is therefore no trivial exercise.

Here is a real-world example from Total Scientific’s own research that nicely illustrates the problem.  Angiography is widely used to visualize the coronary arteries of individuals suspected of having coronary heart disease.   The idea is to identify those at high risk of a heart attack and to guide interventions such as balloon angioplasty, stenting and bypass grafting.   In this respect, the angiogram represents a perfect example of a biomarker composite.  Measures of stenosis in all the major coronary artery regions are to be used to predict a clinical outcome (future heart attack).

At the top level it works well.  Treating the angiogram as a single marker yields useful prediction of future outcome.  Those with coronary artery disease are (unsurprisingly) at much higher risk of heart attack (Figure 1).

Figure 1.  Association between the presence of disease detected by angiography and death following a myocardial infarction (upper table) or death unrelated to cardiovascular disease (lower table).  All data from the MaGiCAD cohort with median follow-up of 4.2 years.

As a useful control, the presence of coronary artery disease is not associated with death from non-cardiovascular causes.  Perhaps the most striking thing about this data, though, is the size of the effect.  People with a significant coronary artery stenosis are only at 3-fold excess of risk of dying from a heart attack in the following four years compared to those with no significant disease by angiography.

Is there more data in the angiogram?  For example, does the total amount of disease or even the location of the lesions provide better prediction of who will go on to suffer a fatal heart attack?  To address this question, we need to treat the angiogram as a collection of separate markers – a measurement of stenosis in each individual coronary artery region.

Among those with some disease, the total amount of atherosclerotic plaque does have some further predictive value (Figure 2).  But again, the most striking observation is the weak nature of the association.  Having a lot of disease versus a little puts you at only marginally greater risk of the fatal heart attack – the total amount of disease cannot be used as a guide as to where intervention is clinically justified.

Figure 2.  Receiver-Operator Characteristic (ROC) curve using total lesion score to predict death as a result of a myocardial infarction (in the “diseased’ group only).  Total lesion volume is better than chance (which would have an AUC of 50%; p=0.011) but carries very little predictive power (a perfect test would have AUC = 100%, and each increment in AUC is exponentially more difficult to achieve).

If the total amount of disease has so little predictive power, does the location of the disease provide a more clinically useful guide?  Previous researchers have attempted to incorporate the location of the lesions into a biomarker composite score.  One example is the Jeopardy Score that assigns weights to disease in different regions of the arterial tree according the proportion of myocardial perfusion that would be lost due to a blockage in that region.  Plaques in proximal locations that cause a greater perfusion deficit ought, in principle, to be more dangerous than stenosis in more distal regions.

Figure 3.  ROC curve using Jeopardy Score to predict death as a result of a myocardial infarction.

Testing this biomarker composite, though, yields disappointing results (Figure 3).  The composite is no better than a simple sum of all the lesions present (compare Figure 2 and Figure 3).   More lesions (wherever they are located) will tend to increase Jeopardy Score, so its unsurprising that Jeopardy Score performs at least as well as the total extent of the disease.  But it is clear that the additional information about the perceived risk of lesions in different portions of the vascular had no further predictive value.

Does this mean that future risk of fatal heart attack is independent of where the lesions are located?  Not necessarily.  The Jeopardy Score biomarker composite was assembled based on a theoretical assessment of risk associated with proximal lesions.  But are proximal lesions really more risky?

Yes and no.  Using the MaGiCAD dataset, we have constructed ‘heat maps’ showing where lesions were most likely to be located among the individuals who died from a heart attack during follow-up, compared with those who did not (Figure 4).  As expected, the left main stem (which feeds both the left anterior descending artery and the circumflex artery) was the site of the most dangerous plaques.  But the next most dangerous location was the distal portion of the circumflex and left anterior descending arteries.

Using this information create a revised Jeopardy Score based on the observed risk in the MaGiCAD dataset now yields a model that significantly improves on the published Jeopardy Score based on theoretical approximation (Figure 4; right panel).  This suggests there really is useful information encoded in the position of the lesions within the arterial tree.

Figure 4.  Left Panel: Heat map of the coronary artery tree showing the relative lesion volume among individuals who died following an MI during follow-up compared to those alive at the end of follow-up.  Dark red represents a 3-fold excess lesion volume among the cases; dark blue represents a 3-fold excess lesion volume among the controls.  Note that the highest risk lesions are located in the left main stem (LMCA), with risk graded from distal to proximal in the left anterior descending (LAD) and circumflex (LCX) arteries, while risk is graded from proximal to distal in the right coronary artery (RCA).  Right Panel: ROC curve using the weightings from the heat map (left panel) to predict death as a result of a myocardial infarction.

Is this the best predictive model you can generate?  Almost certainly not – it turns out that the location of the most dangerous lesions depends on other factors too.  The left main stem is dangerous in younger men (justifying its colloquial designation as the ‘widowmaker’) – but in men over the age of 65 and in women lesions in the left men stem are no more dangerous than those elsewhere in the arterial tree.

Mathematical tools exist to create optimized models combining all these different factors.  One example is the Projection to Latent Structures (or PLS) implemented using SIMCA.  Constructing a PLS model from the MaGiCAD data yields a yet more predictive model (Figure 5; right panel).  Figure 5 illustrates the gradual improvement in the performance of the biomarker composite as more sophisticated algorithms are used to weight the component markers.

All this nicely illustrates how data-driven optimization of biomarker composites can dramatically improve predictive power.  But it does not (yet) give us clinically useful insight.  Because the models have been derived using the MaGiCAD dataset, the ability to predict outcomes in the MaGiCAD cohort (so-called ‘internal predictions’) is likely to be artificially high.  This is particularly true of the PLS model, because PLS is a ‘supervised’ modeling tool (in other words, the algorithm knows the answer it is trying to predict).  Before we can start to use such a biomarker composite clinically, we need to test its ‘generalizability’ – how good it is at predicting death no matter where the angiogram was performed.

Figure 5.  Series of ROC curves demonstrating the improvement in predictive performance with more advanced algorithms for weighting the component markers derived from the angiogram.  Right Panel: ROC curve using the weightings from the PLS model of the MaGiCAD angiography dataset to predict death following a myocardial infarction.

Why might the model not be generalizable?  One obvious reason is that the outcome (death following myocardial infarction) may have been modulated by the intervention of the clinicians who performed the angiography – using the information in the angiogram itself.  It is perfectly possible that distal lesions appear to be the most risky precisely because clinicians perceive proximal lesions to carry the most risk and so treat proximal lesions more aggressively than distal ones.  If that were true, all our heat map would represent is the profile of intervention across the coronary artery tree rather than anything about the underlying biology.  Since patterns of interventions may vary between clinical teams, our highly predictive biomarker composite may apply uniquely to the hospital where the MaGiCAD cohort was recruited.

If this example does not provide all the answers, it should at least provide a list of questions you should ask before adopting published biomarker composites.  Just because a particular composite score has been used in many studies previously you should not assume it represents an optimal (or even a good) combinatorial algorithm.  Usually, combinations are assembled on theoretical (or even ad hoc) grounds and rarely are different combinations considered and compared.

Nor should you assume that combination of component markers will automatically be more powerful than any of the individual markers.  Because the noise in different markers is rarely correlated, but the signal component is more often than not highly correlated, the act of combination inherently reduces power, unless it has been done very carefully.

Before adopting a biomarker composite as an end-point in a clinical trial, you need to understand which components are contributing the greatest noise and which contain the dominant signal.  The results of such an analysis may surprise you.

But most importantly of all, you should recognize that the superficially straight-forward task of combining individual biomarkers is not a task for the uninitiated.  Injudicious combination will reduce rather than increase your power, and even with the most powerful statistical tools available today developing superior biomarker composites is a slow and painstaking task, with no certainty that the composite score that emerges will be much superior to its components.  In short, biomarker composites are more likely to be your problem than your solution.

David Grainger
CBO, Total Scientific Ltd.

The interleukins are an odd family.  One name encompasses dozens of secreted proteins that are linked by function rather than by structure.  And even that common function is very broadly defined: cytokines that communicate between cells of the immune system.

Defined in such a way, its perhaps not surprising that the interleukins have yielded some of the best biomarkers of inflammatory disease conditions, and even more importantly are the target for a growing range of antibody therapeutics.  Interfering with interleukins is to biologicals what GPCRs are to small molecule drugs.

As with GPCRs, though, despite the success of interleukins as biomarkers and drug targets, some members of the superfamily are extensively studied and well understood, while others lie on the periphery largely ignored.  Type interleukin-1 into PubMed and it returns a staggering 54690 papers.  Repeat the exercise for the rest of the interleukins and you make an interesting discovery: although there is a slight downward trend across the family (probably reflecting the decreasing time since each was first described), there are a couple of striking outliers (Figure 1).  Family members who are much less well studied than the rest.   IL-9 has only 451 citations, IL-16 has 414 and IL-20 just 98.

Figure 1 : PubMed Citations for the Interleukin Family in December 2011. Note the log scale.

Are they really less interesting?  Or does this just reflect the positive re-enforcement of previous publications?  Once one paper links a particular interleukin with a disease or physiological process, a crop of papers exploring that link quickly appear, casting in concrete the random process of discovery.  If that’s correct, these unloved interleukins might make excellent targets for research and drug discovery.

Take IL-9 for example: what little is known about this cytokine certainly doesn’t paint a picture of a backwater function undeserving of attention.  IL-9 is a product of CD4+ T cells (probably one of the Th2 group of cytokines that includes the much-studied IL-4 and IL-5) that promotes proliferation and survival of a range of haemopoietic cell types.  It signals through the janus kinases (jaks) to modulate the stat transcription factors (both of which are validated drug targets in inflammatory diseases).  Polymorphisms in IL-9 have been linked to asthma, and in knockout animal studies the gene has been shown to be a determining factor in the development of bronchial hyper-reactivity.

IL-16 looks no less interesting.  It is a little known ligand for the CD4 protein itself (CD4 is one of the most extensively studied proteins in all of biology, playing a key role on helper T cells, as well as acting as the primary receptor for HIV entry).  On T cells, which express the T Cell Receptor (TCR) complex, CD4 acts an important co-stimulatory pathway, recruiting the lck tyrosine kinase (a member of the src family, and itself and interesting drug target being pursued by, among others, the likes of Merck).  But CD4 is also expressed on macrophages, in the absence of the TCR, and here it is ligand-mediated signaling in response to IL-16 that is likely to be the dominant function.

Another interesting feature of IL-16 is the processing it requires for activity.  Like several other cytokines, such as TGF-beta, IL-16 needs to be cleaved to have biological activity.  For IL-16 the convertase is the protease caspase-3, which is the lynchpin of the apoptosis induction cascade, tying together cell death and cell debris clearance.

Like IL-9, polymorphisms in the human IL-16 gene have also been associated with chronic inflammatory diseases, including coronary artery disease and asthma.  But perhaps the most interesting observations relating to IL-16 come from biomarker studies.  Our own studies at Total Scientific in our extensive range of preclinical models of chronic inflammatory diseases have repeatedly found IL-16 to be the best marker of disease activity.   In human studies, too, IL-16 levels in both serum and sputum have been associated with inflammatory status, particularly in asthma and COPD but also in arthritis and IBD.

After years in the backwater, perhaps its time for the ‘ugly ducklings’ of the interleukin family to elbow their way into the limelight.  After all, the rationale for adopting either IL-9 or IL-16 as a diagnostic biomarker, or even as a target for therapeutic intervention, is as good as the case for the better known interleukins.  But the competition is likely to be less intense.

Many years ago, the Nobel laureate Arthur Kornberg, discoverer of DNA polymerase, once said “If, one night, you lose your car keys, look under the lamppost – they may not be there, but it’s the only place you have a chance to find them”.  Sound advice – unless, of course, there are twenty others already searching in the pool of light under the lamppost.  Maybe the twinkle of metal in the moonlight may be your chance to steal a march on the crowd.

Dr. David Grainger
CBO, Total Scientific

They are everywhere: so called ‘present organic pollutants’, or POPs for short.   Since almost all the everyday items that make modern life so much easier emerged from a chemical factory, its not surprising that environmental contamination with organic chemicals is increasing all the time – even ‘environmentally aware’ Western countries.  But maybe it will surprise you to learn they are in your food as well.

New data, published in the Journal of the American Medical Association last week, showed that eating canned soup increased exposure to the compound Bisphenol A (BPA).  Since BPA is a component of many plastics, and is found in lots of food packaging and particularly in cling film, its been known to find its way into food for many years.

In response to the latest study, suggesting that canned food, as well as plastic-wrapped food, can be contaminated with BPA (since modern tin cans, as well as not being made of tin, also have a plastic inner lining), the Food Standards Agency in the UK moved quickly to quell fears:  “Our current advice is that BPA from food contact materials does not represent a risk to consumers” they said.

But is that true?

A British Heart Foundation funded project at the Universities of Exeter and Cambridge have been using the Total Scientific biomarker platform to investigate this question in some detail.  And while the results are not yet conclusive, there is certainly no reason to be complacent.  If the Food Standards Agency had said “There is presently no conclusive evidence that BPA from food contact materials represents a risk to consumers” they would have been correct – but the absence of evidence is certainly not the same thing as the absence of risk.  A more cautious approach is almost certainly warranted.

BPA is an organic compound classified as an ‘endocrine disruptor’: that is, a compound capable of causing dysfunction to hormonally regulated body systems. More than 2.2 million metric tonnes of BPA are produced worldwide each year for use mainly as a constituent monomer in polycarbonate plastics and epoxy resins. Widespread and continuous human exposure to BPA is primarily through food but also through drinking water, dental sealants, dermal exposure and inhalation of household dusts. It is one of the world’s highest production volume compounds and human biomonitoring data indicates that the majority (up to 95%) of the general population is exposed to BPA, evidenced by the presence of measurable concentrations of metabolites in the urine of population representative samples.

In 2008, our collaborator Professor David Melzer in Exeter published the first major epidemiological study to examine the health effects associated with Bisphenol A. They had proposed that higher urinary BPA concentrations would be associated with adverse human health effects, especially in the liver and in relation to insulin, cardiovascular disease and obesity. In their human study higher BPA concentrations were associated with cardiovascular diagnoses (with an Odds Ratio per 1SD increase in BPA concentration  of 1.39, 95% CI 1.18-1.63; p=.001 with full adjustment).  Higher BPA concentrations were also associated with diabetes (OR per 1SD increase in BPA concentration, 1.39;95% CI 1.21-1.60;p<.001) but not with other common diseases.

What that study did not do, however, was determine whether increased exposure to BPA was causing the increase in cardiovascular disease, or was an association due to some confounding factor.

Using our MaGiCAD cohort, these researchers have attempted to replicate these previously published associations, and using the prospective component of MaGiCAD should allow a first indication of whether any observed associations are actually causal.  If exposure to BPA really does increase the risk of heart disease, the implications for safety assessment of BPA and other POPs is significant: we may have to re-evaluate our use of BPA and introduce tighter controls on existing and new chemicals to which people are commonly exposed.

The problem is that it is really difficult to detect a weak, but significant, association between a common exposure and a highly prevalent disease, such as coronary heart disease.  Worse still, because the exposure is so common, even a relatively small increase in risk among those exposed could contribute a significant fraction of the population burden of heart disease, the biggest cause of death in the UK today.  And with every possibility that it is chronic low dose exposure over decades that is responsible for any damaging effects, it is difficult to envision how we could determine whether such POPs are safe enough to justify their use – at least until the harms they cause are detected decades after their widespread adoption.

Indeed, past history shows that chemicals can be very widely used before their harmful effects become known.  The insecticide DDT, or the carcinogenic food dyes such as Butter Yellow are good examples.  It is easy to assume in the 21^st Century that our regulations and controls are good enough to prevent a repeat of these mistakes.

But the emerging data on BPA suggests that this is no time to be complacent.  Just because of the sheer scale of the exposure over so many years, it is far from impossible that BPA has caused more illness and death than any other organic pollutant.

The results from our studies, and other parallel studies by the same researchers, have just been submitted for scientific journals for peer review.  It is only appropriate that the results are released in this way, after rigorous scrutiny by the scientific community (in so far as peer review is ever rigorous).  But those results, when made public, will only add to the concern being expressed about BPA.  There may not yet be a conclusive answer as to the safety of BPA, but it is already time to ask just how much evidence will be needed before it is time to act to reduce our exposure.  Do we need to prove beyond all doubt that it is harmful, or will a “balance of probabilities” verdict suffice?

This is more a question of public policy than epidemiology.  A previous government was willing to ban beef on the bone when the evidence of risk to the population from that route was negligible.  Society needs to make some clear and consistent decisions when to act.  Ban passive smoking, allow cigarettes and alcohol, ban cannabis, allow BPA contamination but ban T-bone steaks.  Sometimes it seems like the decisions made to protect us have very little to do with the evidence at all.

What this study definitely has done, however, is expand still further the range of questions that have been investigated using our biomarker platforms.  Biomarkers may find the bulk of their applications in disease diagnostics and in clinical trials of new therapeutics, but the work on BPA proves that they are also very well suited to complex epidemiological investigations.  Biomarkers, it seems, can do almost everything – except inform the decision about what measures to take in response to the knowledge gained.  Sadly, the politicians are not very good at that either.