The Inflamed Mind

by Edward Bullmore

  If your friend’s doctor can help him to find the causes of his low-grade inflammation he could try to tackle them. He could lose weight if he was obese, which would bring down his cytokine levels. He could try a new dentist or changing his diet. A lot of sensible practices, like physical exercise, sleeping well and avoiding excess alcohol, may have anti-inflammatory benefits. But in terms of lifestyle management this is motherhood and apple pie: very good and very familiar advice that is often very hard to follow. And there may be reasons why he is inflamed that are not so amenable to self-help: how can he escape the stress of caring for a loved one without incurring the guilt of not caring for a loved one? What can he do to change what happened to him as a child? Or what is happening to his body as he gets older?

  In short, what could your friend do differently? And this is why his doctor might have sighed. She would have seen this coming. None of this immunological detective work about possible causes of depression will immediately make much difference to his treatment. There is some evidence that inflamed patients respond less well to anti-depressant treatment with conventional drugs, like SSRIs. So knowing that his CRP is 4.8, and therefore outside the normal range, might make her think twice about prescribing another SSRI if he’s already tried one that didn’t work. That’s not a very exciting therapeutic advance from your friend’s point of view - closing down a treatment option - although conceptually it would be an advance on current practice to use a biomarker like CRP, indeed any biomarker, to predict anti-depressant treatment response. But that’s not much upside. The fact of the matter is that there are no anti-depressant treatments, of any sort, that are focused on reducing inflammation. It couldn’t be different already. There has been plenty of progress in the scientific theory of how the immune and the nervous systems interact; but this new knowledge is not yet enough to make a difference to the real-life experience of depression. The only thing that really drives change in medical practice is new treatment.

  Market failure

  In the decades since Prozac’s launch in 1989, the pharmaceutical and biotech industries have invested billions of dollars in the search for new treatments for depression. The return on that investment - scientifically, therapeutically, commercially - has been discouraging, to put it mildly. Almost nothing has worked. Many promising leads have been pursued, hundreds of clinical trials have been conducted; but there has been no second wave of anti-depressant innovation to follow the wave that started with the accidental discovery of iproniazid and peaked with the advent of SSRIs.

  Acting rationally, companies have stepped back, not wanting to put good money after bad. The level of spending on research and development (R&D) for depression and other mental health disorders has dwindled, projects have been abruptly terminated, scientists have lost their jobs or been reassigned to other therapeutic areas. There are now many fewer new drugs for depression in the R&D pipeline than there were 30 years ago. High levels of investment may not have succeeded in discovering new anti-depressants in the past but it demands gravity-defying optimism to believe that lower levels of investment will be more successful in the future. All other things being equal, less investment will lead to a lower probability of new treatments emerging, and this at a time when depression consistently ranks as one of the single biggest causes of disability among working-age adults worldwide. The level of unmet clinical need could hardly be higher but the level of public and private sector investment is disproportionately low. In an ideal market economy this shouldn’t happen. High levels of demand should stimulate high levels of investment in supplying new products that could meet that demand and close the gap in the market. Money and talent should be pouring into depression research, in theory, but in fact they’re walking away. An economist might diagnose a case of market failure. Industry people tend to say the old business model is broken.

  I witnessed a small piece of this story up close in 2010, when I had been working part-time for GlaxoSmithKline for about five years. One Monday afternoon, I dialled in to an urgently scheduled teleconference to hear the news that GSK was closing its psychiatric research centres in Italy and England, effective immediately. More than 500 people would be made redundant, all ongoing projects would be stopped or spun out into smaller companies, and the Italian site would be sold. We were strategically exiting the whole area of mental health. And GSK wasn’t the only big company making that move; a few weeks later, Astra Zeneca announced an equally swingeing cut to its R&D budget for mental health. It isn’t too hard to grasp the financial logic of these decisions, or to work out why the model was broken.84 It was (and is) more challenging to know what to do next.

  The broken business model for anti-depressant drug development is the way of working that became customary throughout the industry following the path pioneered by Prozac, from about 1990 to 2010. It started with a drug target in the brain, often serotonin, noradrenaline, dopamine or a related molecule. Then thousands of candidate drugs were screened by robots in the lab for their biochemical potency to bind to the target and change the way it worked in a test tube. Once a few candidate drugs had been prioritised from among the thousands initially considered, they were tested on animals, mainly to investigate their safety, and partly in the hope of seeing early signs of efficacy.

  If a mouse is suspended by its tail, so it is hanging upside down in mid-air, it will struggle to get free or reorient itself, for a while; then it will cease struggling and hang quietly. This procedure became known as the tail suspension test and was widely used and described as an animal model of depression for decades, despite its obvious limitations then and now. The principal rationale was that some of the first anti-depressant drugs had sedative side effects, so they made mice struggle less on the tail suspension test, and it was supposed that any new anti-depressant should have similar effects. There was never any convincing demonstration that the mice were depressed, before or after they were suspended upside down. So the industry wasn’t exactly using mice to find new antidepressants; it was using mice to find new drugs that matched the side-effect profile of old anti-depressants. You don’t have to be a Cartesian to see that an upside-down mouse is not a great animal model of human depression.

  The most promising candidate drugs that made it through this pre-clinical process of chemical screening and animal testing were then tested in humans. Phase 1 trials were conducted in healthy volunteers to confirm that the drug was safe and to identify the maximum tolerated dose. Then came the critical step of phase 2: the first clinical trials of the drug in depressed patients. Like the animal testing procedures, phase 2 studies were often designed formulaically to follow a traditional path. A few hundred patients with MDD were recruited and randomly assigned, usually on a 50/50 split, to treatment with a placebo, or treatment with the new drug, for two or three months. At the start and end of the treatment period, patients were interviewed by a psychiatrist, or completed a self-report questionnaire of their depressive symptoms. If the drug-treated patients reported greater improvement in their symptoms than the placebo-treated patients, then the trial was regarded as a success and the drug could move forward to the third and final phase of clinical testing. Phase 3 followed essentially the same experimental plan as phase 2, but on a bigger scale, with studies typically involving thousands rather than hundreds of patients overall. If the placebo-controlled effect of the drug was still statistically significant at the end of phase 3, the data could be submitted for approval of a marketing licence by a government agency.

  The levels of investment increased by roughly an order of magnitude from one million dollars for a study in phase 1, to ten million dollars for a phase 2 study, and 100 million dollars for a phase 3 study. The total cost of getting a molecule all the way to the mental health market was reckoned to be about 850 million dollars in 2010. But most drugs failed somewhere en route. The overall probability of success was less than 10% and the few drugs that made it all the way had to generate enormous amounts of money if they were to recover their own d
evelopment costs as well as the sunk costs of all the failed trials of less successful candidates. The only happy ending was a commercial blockbuster, a drug that was capable of earning billions of dollars a year, by being prescribed to the largest possible number of depressed people.

  Looking back, it is not surprising that this model eventually broke. It seems more remarkable that there was ever a time when it was not broken. Nothing about it now makes sense scientifically. The choice of targets and animal tests was often designed to prioritise product line extensions, or “me too” drugs that were as close as possible to successful precursors. To put it bluntly, the industry kept hammering away at serotonin, dopamine and related targets rather than successfully exploring alternative targets for more innovative drugs. And there was generally a “one size fits all” approach to clinical trials and marketing: an anti-depressant drug was assumed to be equally good for all patients with depression. Not much effort was usually made to understand how a physical agent like a drug could have effects on a mental state like depression. Old-school clinical trials didn’t measure biomarkers, or sequence DNA, or look at brain scans. To be fair, not all these biomedical techniques were available in the 1990s and early 2000s. And some of the methods that might have been very useful then, like a brain scan for serotonin levels, are still not available today. But the absence of any biological data from the industry trials contributed to a lack of deep understanding about how the drugs worked or which patients they might work best for: as I later discovered in my Molière moment at the Maudsley Hospital.

  The business of the blockbuster model was not too troubled by the philosophical or farcical paradox of an anti-depressant drug in a Cartesian world. If a development programme could leap from the tail suspension test to positive phase 3 data then it was legitimised by its own improbable success. What more do you want? But once the market was crowded by the early winners of this lottery, and commercial success had become increasingly elusive for the “fast followers”, it was the model itself that was found wanting. It couldn’t explain its failures. It couldn’t predict its successes. It was scientifically exhausted and it went out of business. That much was not a market failure; it was the inevitable fate of a business model that has run out of steam and been overtaken by market forces. The same thing happened about 150 years earlier to the once-thriving business of herbal remedies for unbalanced humours.

  Economists like to talk about creative destruction - the death of old businesses at the hands of the market creating space for new and better businesses to flourish. Sometimes the old business may be disrupted by an insurgent new business competitor, and sometimes the old business can collapse for other reasons, before a strong competitive business has been marketed. When the old business model for anti-depressant drug discovery collapsed, in 2010, it wasn’t because a new business model for depression was ready to replace it. It was simply because the old business model was unsustainable: not enough return on investment; not enough therapeutic innovation to justify the massive development costs. The old model died before a new model was ready. And there is no law of economics according to which a new model must be born within six months, or six years, or 60 years, of the death of the old model. It could take any amount of time for a company or a sector to reinvest in an area of recent business collapse. The economic destruction may be necessary but it is not always rapidly creative.

  A few weeks after that Monday afternoon teleconference, I asked my boss at GSK if he thought the company would ever reinvest in depression and psychiatry. “I’d never say never. But if we were going to go back there,” he said, as gravely as if the destination were Chernobyl, “it would have to be completely different. What we’re not going to do is stop, wait a bit, and then start doing exactly the same thing we did before all over again. So don’t ask me for tens of millions to jump back into old-school phase 2 because that’s not going to happen any time soon. First, you’ve got to be able to tell me how it’s going to be different next time.”

  Beyond blockbusters: better but not bigger than Prozac

  There can be no more talk of panaceas. We will need to leave behind the idea that depression is all one thing, in much the same way that we no longer think of cancer as one multiheaded monster disease but as a collection of thousands of different kinds of diseases. We will need to recognise that there could be many different causes of depression and therefore to challenge the possibility of a panacea. How could any single treatment, an SSRI or a course of cognitive behavioural therapy, say, conceivably provide the best possible treatment for all patients, regardless of the many different underlying causes of their depression?

  Panaceas are scientifically ruled out. We will have to think instead about how to identify major causes of depression, and how to define groups of depressed patients who share a common cause, and might particularly benefit from a specific treatment. This approach is obviously good from the patient’s perspective because it reduces exposure to the risks of treatment for those who are least likely to benefit. Translating the science of neuro-immunology to anti-depressants along these lines, we will need to design treatments that are targeted at inflammatory mechanisms that cause depression in a subset of depressed patients, not necessarily all patients with depression. The expectation is that anti-inflammatory drugs could work well for patients with inflamed depression. There will be other patients with depressive symptoms, who are not inflamed, who are likely to get the most benefit from existing anti-depressant treatments or any new non-immunological treatments that might be developed in future.

  Prozac and its cousins were blockbusters in two ways: massive commercial success and virtually unbounded licence. They were used as if they were panaceas, to provide a one-size-fits all treatment for depression (and many other disorders). The next generation of anti-depressant drugs will probably be more personalised products that can offer a major therapeutic benefit to patients who are depressed for a particular reason. The development and launch of new anti-depressant drugs is likely to be coupled to so-called companion diagnostics, biomarkers that are validated and licensed for use in conjunction with the new drug. A simple clinical procedure - perhaps a blood test - will be used to predict which depressed patients are most likely to benefit from the new drug. Could this kind of niche-buster product be commercially successful, on the same scale as an old-school blockbuster anti-depressant like Prozac?

  Who knows? But from a commercial perspective, the size of the potential market is obviously an important consideration. How big would the market be for a drug that worked very well as an anti-depressant but only worked for the percentage of patients whose depression was linked to inflammation? That is going to depend on the cut-off criterion used to define which patients are inflamed and which are not. It is also going to depend on whether you’re focused only on depression as a mental illness, typified by the conventional psychiatric diagnosis of MDD, or whether you’re also prepared to think about co-morbid depression in patients, like Mrs P, who have a physical illness. To get a rough idea, let’s start from the fact that about 350 million people, or about 7% of the world’s population, had an episode of MDD in 2012. How many of these people would we expect to pass a blood test for inflammation? If we used CRP as the biomarker, and 3 mg/L as the cut-off point for being inflamed, then we might expect about a third of the patients with MDD to be eligible for treatment with a new anti-inflammatory drug. That’s more than 100 million people. One of the few good things about there being an enormous number of depressed people in the world is that commercially it creates an opportunity to provide a personalised product, for a biomarker-defined niche in the market, with mass-market economies of scale.

  New drug development for inflamed depression will mean validating the blood tests that will be used to decide if depressed patients are eligible for treatment; and then testing the drug in groups of patients that are predicted by the inflammatory biomarker to be responsive (or not responsive) to treatment. This will all take a great deal of time an
d money to do rigorously to the standards needed for the licensing of a new anti-depressant medicine. In the best case, I think, new anti-inflammatory drugs might become available to patients, like Mrs P, with co-morbid depression about five years from now; and for some patients with MDD about 5-10 years from now. That might seem like a long time. But from a hardnosed industry perspective the probability of success is still less than 50%. Remember, most drug development projects fail, especially anti-depressant drug development projects, where the percentage of clinical trials reporting positive outcomes has been dismally low at times. If you asked around the pharma and biotech industries now, I think you’d get the overall impression that there might be a 20% chance that anti-inflammatory drugs could really make it all the way as a new class of anti-depressants. Positive new clinical trial results in the next few years could increase that 20% probability of success dramatically. But there have been many false dawns in the history of psychiatry and the recent surge of optimism about immunological anti-depressants could yet prove to be another one: there’s still risk in the package, as they say. And there will be until there is convincingly positive clinical trial data - which we haven’t yet seen.

  Encouraging the industry to move forward into clinical trials, there are dozens of anti-inflammatory drugs that have already been developed or licensed for other disorders that could potentially be useful for treatment of inflamed depression. The industry jargon for this is repurposing.85 It makes it possible, in principle, to target a range of immune mechanisms for depression without incurring the costs associated with developing a new anti-inflammatory drug from scratch, from the first biochemical screening studies through testing in animal models to phase 1 safety studies. By allowing companies to go straight to phase 2, repurposing can make it a lot cheaper, less time-consuming and less risky to discover whether a drug that is already known to hit its target safely in the human immune system could be effective in depressed patients.