This is the second of two parts on the future of health care over the next 25 years, which I developed from a series of different, but related, presentations I made to a wide assortment of organizations. The first part, which dealt with what such a system might be like, was published on March 5th. This part covers how we might get from here to there.
There are three principal issues that will drive health care over the next quarter century: an aging population; technology; and money. They are linked, but different. Let’s start with demographics.
For the past 20 years or so we’ve lived in a period of relative demographic calm because most of the key generations have stayed within a single phase of their life cycle. This is about to change, as three of the key demographic cohorts are about to undergo major demographic transitions. The first group is mature adults (those born 1938 or before, currently in their 70s and up), who have been described as the “oldest elderly.” A steadily rising percentage of this group are now moving into the stage where they can no longer manage their lives and affairs without assistance.
There will be a dramatic increase in demand for services for this group, especially in chronic care facilities, as this group is growing faster, in percentage terms, than any other age group in the population. Moreover, the children of this group are the boomers, and we boomers are not known for our sweet reasonableness. This means boomer children will demand government help in coping with their parents. The net result is going to be a steadily increasing demand for help with the oldest elderly in our population.
Trouble ahead for boomers – and everyone else
Next are the baby boomers themselves (born 1947 to 1967, turning 43 at the low end and 63 at the high end). In 10 years’ time, they’ll be between 53 and 73. This means the leading edge boomers are now entering the transition to retirement. By my estimates, the annual cost of health care per person remains relatively level until about age 55, at which point it starts going up almost exponentially. The boomers, the largest generation in history, are now entering the high rent district of health care, and although I’ve seen reports that claim that health care costs are going up because of the increasing costs of treatment, and not because of an aging population, I flatly don’t believe it. As one example, the Government of Ontario, in a 2005 publication, projected that health care spending there will rise to more than 55% of program spending (i.e., total spending excluding debt service) by fiscal 2024-25, and I believe this underestimates the real situation as these estimates were made before the recent recession. By 2035, I expect Ontario to be devoting something in excess of 60% of its program spending to health care – and that’s typical of jurisdictions throughout the U.S. and Canada. This will create all kinds of fiscal and financial problems, as well as acrimonious debate at state and provincial governments across North America – and, indeed, throughout the developed world. These costs may even bankrupt all major governments in the developed world, because these programs are essentially pay-as-you-go, and demographic support for the biggest generation in history is simply not there. If governments are to avoid bankruptcy, then people in the developed world are not going to get the health care we want and expect.
Finally, there are the children of the baby boomers, called the “echo boomers” or “echoes”, born roughly between 1977 and 1997 (currently between 12 and 32). In 10 years time, they will range between finishing their formal schooling and approaching their 40s. The echoes are the Next Big Thing for any organization that hires and employs people, for they are not only the next generation of clients and customers, they are also the next generation of health care professionals. The problem is that there aren’t as many echoes as there are boomers, with the result that we are going to experience an increasing shortage of health care professionals for the next 10 years. It will improve slightly after that, but then get substantially worse beyond 20 years, especially in Canada, unless Canada and the U.S. import and certify substantially more foreign doctors, or unless technology provides new solutions.
So where are all the customized drugs?
Next let’s look at the very real promise, and the phony pursuit of personalized drugs. The pharmaceutical industry is talking a good game when it comes to personalized drugs, but I don’t think they mean it. If you look at the current state of such drugs, the industry tends to point to two drugs in particular: Herceptin and Gleevec.
Herceptin (from Genentech) is the exception that proves the rule. It is much talked about, but with few exceptions, its example hasn’t been much followed by pharma companies. In my opinion, the reason why Herceptin is an exception is that the screening test that accompanies it eliminates roughly 65% of potential patients, producing a much smaller potential market. Since big pharma companies are largely run by marketing people and bean counters, they hate the idea of cutting down on potential markets. They would much rather find another blockbuster, even if it means discarding a perfectly effective drug that doesn’t have a large-scale market. And yet, since the overall percentage response rate in the overall potential population for Herceptin is less than 10%, Herceptin would probably not have been approved without a genetic screen. With the screen, a failed drug was rescued, serves as a niche drug, and still produces $1 billion in revenues per year. Yet, despite Herceptin’s success, as far as I know, Genetech doesn’t seem to have made any effort to improve the focus of the diagnostic test; it’s seen as “good enough,” even though they could increase the efficacy of the drug by narrowing the focus further.
By comparison, Gleevec (from Novartis) seems like the industry’s ideal – but there’s a catch. Gleevec targets a particular protein in chronic myeloid leukemia (CML), and produces a 75-80% response rate, which is head & shoulders above competitive drugs. This is the industry’s ideal, but it was targeted on that protein because it was already known there was a relationship between that protein and CML, which is the reverse of most drug research.
Pharmaceutical companies dragging their feet
There are probably other “personalized drugs” in development, but nowhere near the number you would expect from Herceptin’s success as a $1 billion a year niche product. I believe this is because the pharma industry is dragging its feet. And I further believe this is a mistake: pharmaceutical companies should be eager to pursue this approach, as it will make them more money than ever before. The problem is that it requires a different business model with which they aren’t familiar. All they know is that the marketing model they have followed for decades is based on big blockbuster drugs that sell billions of dollars of drugs to millions of people. They don’t see how selling millions of dollars to thousands of people can replace that – yet it can, as I’ll come to in a moment.
Moreover, in both cases, Herceptin and Gleevec involve a single protein, a single test, or a single genetic marker, yet most genetic indicators will be more complex than that. Mostly there will be multiple indications, and a multivariate approach will be called for, which has been a problem in drug research so far.
According to Tufts University, in 2006 it cost $1.2 billion to develop a new drug and bring it to market. Most of that is the opportunity cost of selecting one compound for development compared to other possibilities, and investing the 15 years necessary to bring it fruition. And a hefty chunk of the cost associated with drug development comes in drugs that get as far as Phase II or III clinical trials, and then fail because they either are not effective enough, or they have unacceptable side-effects.
But if you can identify the populations for whom such drugs either have low efficacy or unacceptable side effects, you can still market such drugs profitably, albeit to much smaller segments of the target market. So, which is better? To get nothing from a drug, and flush all of the associated costs? Or to target it to a smaller population, derive lower revenues, but at higher profit margins? To date, the major pharmaceutical companies have generally decided they’d rather flush such drugs than take them to market. If your revenues are $20 billion, and Wall Street expects 20% growth, you have to come up with an additional $4 billion in revenues every year. With that as your target, a drug that might produce $100 million in annual sales is not worth pursuing. Yet, according to one research group with whom I’ve worked, “A failed drug is an expensive lost opportunity.”
Improving the beauty contest in drugs
A major pharma company may have 300 compounds it is considering at any one time, and must pick and choose which ones it will develop. This becomes a matter of internal politics, and a bit of a beauty contest. If, instead, the pharma companies developed companion diagnostics based on genetic indicators right from the start, they would have a much better way of predicting the potential market for each drug, and hence its ultimate profitability. The direct marketing industry knows this well. Broadcast advertising – TV and newspaper ads – sells on the basis of cost-per-thousand. Direct marketing, like direct mail, sells on the basis of cost-per-sale. And successful Google Adwords marketers and the like measure success on the basis of cost-per-dollar-of-profit. Companion diagnostics are a way of coming closer to being able to predict cost-per-dollar-of-profit, and that should be the yardstick pharmaceutical companies use going forward. Moreover, it’s a yardstick that would lead them to substantially bigger profits and lower costs.
Which brings me to new research tools. Although statisticians deny it vehemently and for obvious reasons), statistical analysis is no longer sufficient in a world where relevant data are growing by orders of magnitude, and multivariate analysis is required. The old tools just aren’t up to the task, yet companies cling to them because, as Abraham Maslow once remarked, “When all you have is a hammer, you tend to see all your problems as nails.” Yet, new tools are emerging in strange and unusual places, such as mathematical models of fuel combustion in jet engines, and software techniques for designing radio antennae, none of which have any credibility with pharma companies – but which can solve these kinds of massive data problems. By using such tools, and embracing rather than avoiding pharmacogenomics, I believe pharmaceutical companies can reduce drug development costs, perhaps by an order of magnitude, produce more successful drugs, albeit for smaller potential markets, and sell them for higher profit margins because of their higher efficacy. Let’s look at one potential new tool set to see if this is realistic.
Genetic programming – a new tool set
I’m going to describe one specific tool that’s already emerging into the marketplace, and about which I – accidentally – know quite a bit, called genetic programming, or “GP.” Genetic programming is a machine-learning technique where the system evolves by reinforcing success. It uses the idea of natural selection to discover solutions. The solutions that work best are combined to discover even better hybrids, much as cross-breeding horses can create offspring that are faster and more robust than their parents. GP is not an artificial intelligence system, and there is no attempt to mimic human reasoning. GP’s advantages are that it:
- Solves the problem of too much data.
- Integrates large & diverse data sets.
- Facilitates the unbiased discovery of key factors, especially where the key factors are unknown.
- Ignores factors that turn out to be irrelevant or unimportant; and
- Creates human readable models that can serve as indicators for future research.
Moreover, GP is a non-linear method, which means it can be dramatically faster than conventional analysis techniques, especially for multivariate analysis involving large amounts of data. Indeed, large amounts of data tend to produce more robust results.
I’m about to describe a company called Genetics*Squared, but in the interests of disclosure, I must also say that I know the principals, I have worked with this company, and I own shares in it. This is also how I know so much about it. However, since Genetics*Squared (or “G*2”) is privately held, there are no shares available to outsiders anyway, so the disclaimer is academic.
G*2 is a “dry” biotech company that uses GP to analyze clinical trial data in order to produce prognostics and diagnostics predicting who will respond to a given therapy or pharmaceutical. G*2 is about to seek FDA approval for a prognostic test for colorectal cancer to determine which patients, following surgery for colorectal cancer (CRC), are at risk for a relapse and should seek additional treatment. Preliminary indications are that the test they developed has 85% sensitivity, and 90% specificity. These results compare with the National Comprehensive Cancer Network (NCCN) guidelines of 73% and 38%. This means G*2’s test is very good at identifying the people who are – and are not – at risk for recurring colorectal cancer. The test does not suggest the form that treatment should take; that’s up to the patient and her doctor to explore and decide.
A logical next step from there might be to use genetic programming to run clinical trials with those patients who are at-risk for a relapse in CRC to see which ones would respond to an off-patent, generic chemotherapy drug, such as 5-FU (Fluorouracil), a chemotherapy drug that’s been around since the 1950s. For those identified at high risk for CRC recurrence, and for whom 5-FU is found to be effective, as identified by such a diagnostic, the efficacy and cost of 5-FU would have the potential to completely undermine the market for any new drug for whom there was no screening test.
For instance, suppose G*2‘s test were to eliminate 66% of a potential market for a new chemotherapy drug for CRC, and suppose further that one-third of those at risk for relapse could use generic 5-FU instead of the newly developed drug. This means that instead of a potential market of 100% of those people who have had surgery for colorectal cancer, the pharmaceutical company would have a potential market of only 22% of such patients. (Please note that I’m making these numbers up purely for illustration purposes.) Hence, using more sophisticated screening, and a low-cost, generic drug might not only be more effective, but could spoil the potential market for a new drug that does not identify the appropriate niche for its use. It is this low-cost competition that will ultimately drive the pharma industry toward pharmacogenomics, not regulation: the threat that smaller companies could destroy the market for blockbuster drugs. Moreover, how many failed drugs could be resurrected with an appropriate genetic screening? And how much would that save in R&D costs? It will ultimately be money that will make pharmacogenomics appealing.
On beyond genetics
Beyond genetics, IT will also force a reshaping of health care. Moore’s Law, coined by Gordon Moore of Intel fame, states that computers will double in speed, and halve in price every 18 months. (Actually, Moore’s Law refers to the number of transistors on a chip, but this is the effect of it.) Yet Moore’s Law, even though it represents exponential growth, is too conservative. Not only is the pace of change accelerating, but the rate of acceleration is increasing. This implies that a computer for a given price will be roughly 1000 times faster in 10 years’ time than it is today. This is going to produce changes over the next 10 years that are at least twice as dramatic as the changes of the last 10 years. And technology is going to have direct and important changes in medical practice.
Let’s start with the basic building block of the global computer network that tracks and identifies new diseases and epidemics that I described earlier. The precursor to the computer companion, which forms that basic building block, is already in widespread existence today: the smartphone, such as the BlackBerry or iPhone. For instance, I understand that there is an application for the iPhone that communicates with a heartbeat monitor to monitor your health heartbeat-by-heartbeat, much as I described it earlier. Only this isn’t science fiction; this is today’s reality. As smartphones become more powerful, more globally connected, and work with devices that monitor and analyze different aspects of your health, I expect they will gradually morph into all-purpose assistants, including health monitors.
When coupled with the intricate knowledge of our individual genomes, and the dramatic increases in our understanding of what the various markers in our genomes mean, these computer companions will become incredibly powerful and useful agents on our behalf. And analysis of individual genomes is not that far off. The first reading of a human genome was the focus of the Human Genome Project, and it cost billions of dollars. Today, the cost of reading a genome is approaching $1,000, and within five years, or ten at the outside, it will approach $100 per person. And once read, your genome will provide a continuing source of new insights into why your health is the way it is, plus why your body is the way it is, as we learn more about how to interpret the genetic markers buried in each genome.
The greatest medical tool humanity has ever had
When coupled with global analysis of lifestyle factors, food, exposure to particular viruses, bacteria, and chemicals in our personal environments, we will begin to build a broad understanding of what helps – and hurts – us, both as individuals, and as a species. This will be, I believe, the greatest medical tool humanity has ever had.
But it won’t end there. There will be lots of developments in lots of areas. These range from dramatically improved x-rays, CAT scans, and imaging of all kinds, to computer-oriented solutions to difficult problems, to the rise of robots, both for surgery and for attending to pateints.
Let’s start with an example in imaging. Dr. Otto Zhou of the University of North Carolina is working with Siemens to bring out a much clearer and more precise x-ray technology based on electron-field emission[i].
This allows for an array of receivers so that a CAT scan, for example, can be taken all at once instead of progressively, so that the images will be dramatically sharper and more detailed. It also allows for imaging at the same time as treatment, so that health care professionals can watch precise images as the same time as they zap a cancer tumor, for instance.
Up until now, robots have been seen as science fiction phantoms, good for Hollywood, and not much else. All of the robots to date are seen either as being bolted to the floor in car factories, or as cute little toys that don’t do much. But did you know that Toyota is pouring billions of yen into producing household robots, and plans to start marketing them in their home market, in Japan, this year?
Over the next 10 years, we are going to see the gradual emergence of much more flexible and powerful robots. BigDog & HRP-4C are both examples of today’s robots. They are awkward and clumsy. Rather like human toddlers, they are taking their first steps, and those first steps are ungainly and tentative. Yet, like human toddlers, they will improve rapidly, and this will be powered both by dramatically faster computers, and increasingly sophisticated tools and techniques to use them.
According to Raymond Kurzweil, an inventor and technologist, by 2016, a $1000 computer will have more computing power (not “intelligence”) than a human brain. By 2036, he projects that a $1000 computer will have more computing power than the entire human race. And increasingly sophisticated software will harness this power into seeming-intelligence that will first rival, and then exceed our own. And in health care, this rapidly accelerating intelligence will produce a rapidly increasing flow of new discoveries and treatments.
It’s not all good; trouble ahead
Finally, let me address one of the problems ahead, just to demonstrate that I’m not a pollyanna, and that I do appreciate that the road to dramatically improved healthcare will be rough. The biggest problem ahead is summarized in two words: money and politics.
We know that money spent in preventing disease is generally more effective than curing it. Hence, promoting the use of condoms for safe sex is more effective than trying to deal with an AIDS epidemic, but politics and social attitudes often get in the way. Moreover, we know that some choices make no logical sense. We know that we spend, on average, about half of all the money we will ever spend on health care in the last six months of our lives. So sensibly, we should know when to say enough is enough. But paranoia about litigation, coupled with an attitude of preserving life at any cost, seems to prevent us from even discussing such issues, particularly when it’s your mother, your wife, or your son that’s involved.
Yet, I don’t think I need to convince anyone here that our health care systems in both America and Canada are leading towards bankruptcy. The U.S. Federal Reserve and the Government Accounting Office essentially agree on the extent of the projected unfunded liabilities of the U.S. federal government. The Fed estimates it at $99.9 trillion, while the GAO estimates it at $101 trillion – and almost 2/3 of these estimates relate to health care costs.
And while the Canada Pension Plan is in much better shape than Social Security in the U.S., Canada is in no in better shape in terms of its promises on health care. The IMF published a report, which has been refashioned as a book by Peter Heller and published by the IMF, called “Who Will Pay?” In this book, Heller quotes the IMF as indicating that as of 2002, the U.S. had explicit net debt of about 45% of GDP, and Canada had explicit net debt of about the same (Canada’s has moved down since then, while America’s has shot up). However, when you add the implicit debt due to future promises, in 2002, America’s explicit plus implicit debt amounted to about 265% of GDP, and Canada’s amounted to about 415%. What is beyond dispute is that both governments – indeed, all governments of all developed countries – are going to be under enormous financial pressure because of the aging of their populations. What we don’t know – and what I would strongly suggest you should prepare for – is how governments and voters will cope with being unable to give boomers the health care they demand. To attempt to do so is to court being voted out in favor of some demagogue who promises the impossible.
And there’s another aspect to the future: the war against infectious disease has not yet been won. We are already experiencing a silent pandemic about which virtually nothing is said: the spreading of antibiotic-resistant superbugs. And even here, policy choices are clouded by what should be extraneous issues, such as creating deliberate breeding grounds for superbugs by feeding antibiotics to livestock, a practice advocated and protected by the farm lobby. As well, neither governments nor pharmaceutical companies want to finance the development of new antibiotics. Governments don’t want to do it, because it looks, to a bean counter, like an extra expense. Pharmas don’t want to do it because there’s no guarantee of a large enough market. Moreover, we have no true cures for viruses at all, except vaccination ahead of time, and there is not much money going into research in this field, either. Yet, in the war against infectious diseases, I can promise you that we have not won the war, only an opening skirmish. We have pitched battles yet to come, whether in the emergence of a viable airborne ebola virus, or some completely unknown bacteria infection.
For example, consider the H1N1 flu pandemic. It had the potential to be as bad as the Spanish flu of 1918-20, although we also have a better idea how to contain it. But now imagine the emergence of a hyperbug – a strain of bacteria resistant to all known antibiotics – emerging in the midst of the flu pandemic. This is not unlikely; overcrowded hospitals and over-tired health care workers create a likely breeding ground for just such a hyperbug. The day will come when we can analyze and respond to such threats in a matter of days, but we’re not there yet, and millions of people will die until we get there.
Two questions everyone wants answered, but no one wants to ask
At the core of the health care debate, there are two questions that no one wants to ask, but that everyone wants answered. First: How can I get someone else to pay for my health care? And second: Do we want a more effective & more expensive health care system, or a less effective & less expensive health care system? And the problem is that voters (and hence politicians) want a more effective, less expensive system, which is not one of the choices. Worse yet, we are not even discussing these issues. Instead, we are shouting at each other about useless matters that won’t solve the problem.
Change will come slowly at first. Social systems have tremendous inertia. Drug approvals are slow. Payers are slow to accept new, more expensive, more effective drugs. Doctors are so swamped that they are reluctant to take the time to learn new fields, like genetics, that impinge on their own, so that practice in the field can often take years or even decades to catch up with research in the lab. And the general public, and the politicians they lead, don’t take the time to understand the debate, with the result that they fall prey to demagogues who cloud the issues for their own reasons. In Canada, it’s a phony “public vs. private” debate. In America, it’s the misleading “socialized health care is worse than dying” debate. Both debates are largely irrelevant to the central issue of financing health care. Worse, they divert attention and vital resources from the real issue.
Changes will start slowly. But the pressure from an aging population, an Internet-informed voting public, and the startling developments from science will start to move the log jam, slowly at first, and then faster and faster. If you are not prepared when the pace picks up, you will quickly be left behind. I wish you good luck, and God speed. Thank you.
[i] “Another look inside: The way medical X-rays are generated is over 100 years old. Time to update it”, The Economist, July 30, 2009.