Case Study: Diabetes
Just as everyone knows that too little serotonin causes depression, everyone knows what is wrong with you if you have diabetes and what should be done about it. Diabetes is popularly known as “sugar” in some communities, and that’s accurate because the basic problem in diabetes is a too high blood sugar level. This is true in both sorts of diabetes, type 1 (juvenile onset) and type 2 (adult onset). The difference is that in type 1, the problem is too little insulin, and no treatment is possible without injecting insulin. In type 2, however, some of the problem is resistance to insulin in the body’s cells, and this can sometimes be treated with oral medications, without the need for insulin injections. But in any type of diabetes, the mainstay of successful treatment is to restore the blood sugar to a normal or at least near-normal level.
If anyone doubted this narrative about the cause and treatment of diabetes, their doubts would be removed by watching any of the advertisements on daytime TV aimed at the US Medicare population with type 2 diabetes. These ads make clear that it is an absolute necessity that anyone with the diagnosis of type 2 diabetes possesses and frequently uses a home glucose monitor. Your physician would, of course, tell you this (and, by implication, any physician who does not is not worth listening to). And don’t worry about the finances, because Medicare will happily pay for most of the cost of both the monitor and the test strips. (Indeed, the device companies find that selling the test strips is such a lucrative business that they will sometimes give away the monitor for free.)
Now, watching these ads, one would never know that when randomized trials are conducted, in which half the subjects with type 2 diabetes are given monitors and the other half not, no difference is found in either control of blood sugar or other health outcomes (Farmer et al. 2012: e486). Nor would one know that a number of major studies over several decades have shown with a high level of confidence that normalizing blood sugar in type 2 diabetes is not a particularly good strategy for achieving the most important health goals (Boussageon et al. 2011: d4169).
Why is diabetes bad for you? It is true that once a certain very high level is reached, the hyperglycemic state itself causes one to feel bad and function poorly, so at that level, it does in fact make a real health difference to lower blood sugar. Many people with mild to moderate type 2 diabetes, however, seldom if ever reach that level of hyperglycemia.
The much more serious problem with type 2 disease is the complications. People with diabetes have a much higher rate of serious and potentially fatal complications, including heart attacks, stroke, kidney failure, blindness, and blood vessel disease requiring limb amputation. So it’s reasonable to ask what the scientific evidence shows about the relationship between strict control of blood sugar and preventing these serious complications.
The largest and most extensive study to address this question was the UK Prospective Diabetes Study (UKPDS), which enrolled a large number of patients and followed them for ten years, a feat unlikely ever to be repeated. The main UKPDS investigators confidently expected to show a close correlation between blood sugar control and reduced rate of complications, and so that’s how they reported their results, even though they had to play fast and loose with a number of statistical analyses to get that outcome (UK Prospective Diabetes Study Group
- 2008) . Others, looking at the UKPDS body of data more critically, have pointed out that the data simply do not support that conclusion (Montori and Fernandez-Balsells
- 2009) . Particularly with regard to the complications that involve the larger blood vessels (heart attack, stroke, amputation), there was no good evidence that reducing the blood sugar lowered the rate of complications. One medicine, metformin, turned out to be very good at preventing bad complications. Other medicines, however, that lowered sugar as much or more had no such preventive effect, so it seemed that metformin exercised its benefits in a way separable from its glucose-lowering properties.
If one looks at UKPDS in this critical manner, it seems a good example of why association needs to be distinguished carefully from causation. There is a good deal of evidence from epidemiologic studies that there’s a close association between elevated glycohemoglobin (the definitive blood test that shows that one has had a high blood sugar over an extended time period) and higher rates of complications from diabetes. It would then seem eminently logical that treatment that lowers gly- cohemoglobin level (by better controlling blood sugar) would reduce the rate of complications. But evidence doesn’t seem to support this conclusion. Since elevated glycohemoglobin is a marker for more severe diabetes, it seems that those with more severe diabetes are more likely to suffer complications and vice versa. But drugs that lower blood sugar generally seem ineffective in turning severe cases into mild cases. By contrast, attempting to control diabetes through diet and exercise seems more promising in terms of reducing complications, but this is not the place to discuss that.
In light of the lessons we have learned from a closer scrutiny of the UKPDS’ data, we’d naturally expect that when other studies used medications to lower blood sugar levels in type 2 diabetes, the same outcome would occur. And to date, this is precisely what these further studies have shown. In general, studies that use medication to more strictly control blood sugar do not significantly reduce the rate of complications. However, such a regimen almost always causes harm, because the strict-control patients suffer more frequent episodes of low blood sugar (hypoglycemia). And hypoglycemia can be very serious, leading to seizures and coma; and the risks of and damage from hypoglycemia increase as one gets older.
What’s striking about these further studies is that diabetes experts, who might be expected to say, “Oh well, that’s what UKPDS suggested, so no big surprise here,” act astounded each time such a study is reported (Kishore et al. 2012). It seems that the comfortable narrative of diabetes = sugar is so deeply entrenched in both medical and popular thinking that even physician-scientists cannot quite let go of it, no matter how much contrary evidence accumulates.
Medical historian Jeremy Greene has studied where this compelling narrative comes from in his book, Prescribing by Numbers: Drugs and the Definition of Disease (Greene 2008). During the middle of the twentieth century, beginning with the famous Framingham Study of heart disease risk, medical science happened upon a new concept—the idea of the risk factor. In older times, either you were healthy and the doctor left you alone or you were sick and the doctor treated you. Now suddenly there was a new way that you could be healthy and yet need the doctor’s care. You might have a risk factor which, if unattended to, made it more likely that you’d suffer some really serious illness later on but that could be mitigated with the proper medical attention, including drugs.
Everyone at the time thought that risk-factor medicine was a marvelous advance. A common criticism of the older style of medicine was that it failed to attend to prevention. Risk-factor medicine, by contrast, seemed to be preventive medicine par excellence. What’s not to like?
As Greene reviews the history, what failed to get noticed at this time (or that more accurately, seems meaningful only in hindsight) was the very close cooperation between the pharmaceutical industry and medicine in making the turn toward risk-factor interventions. What the industry most wanted, and medical science appeared eager to supply, was “prescribing by the numbers”—developing and encouraging the widespread use of simple blood tests to identify a threshold level of a risk factor and then persuading doctors that a healthy patient who had such an abnormal lab result ought immediately to be placed on a medicine, ideally, an expensive medicine that needed to be taken for the rest of one’s life.
“Prescribing by the numbers” is very near a dream world for a pharmaceutical marketer. According to the recently adopted mantra of patient-centered care, the ideal research study randomizes subjects to the drug or no-drug condition and then follows them long enough to see how many develop an outcome that actually matters in people’s lives, such as death or a condition like a heart attack that causes significant disability and hospitalization (Washington and Lipstein 2011). Such a study (like UKPDS) takes a long time and is very expensive. Moreover, it is hard to predict what will be the actual outcome of such a study; and drug marketers hate that uncertainty.
Consider by contrast the sorts of studies drug companies love, aimed at what evidence-based experts call “surrogate markers.” Once you have decided that high blood pressure or high blood sugar indicates conditions that count as risk factors for later heart disease, you then do a study to show that a drug lowers blood pressure or blood sugar. Typically a drug will show an effect on the surrogate marker within a few weeks, making for a much shorter and hence cheaper study. Based on preliminary data, it’s much easier to predict a positive outcome involving surrogate markers than it is if one were studying actual clinically relevant outcomes. And, there’s a bonus: It’s not uncommon that a drug will lower a surrogate marker promptly but show a serious adverse reaction only if taken for some months. If the company is lucky, it can get in and out with its study in time to “prove” success with the surrogate marker, but not run the study long enough to pick up any unwelcome signs of bad side effects.
The recent career of rosiglitazone (Avandia) nicely illustrates the success of “prescribing by the numbers.” Rosiglitazone is one of a class of drugs introduced with much fanfare for type 2 diabetes in the late 1990s, because they promised a new biochemical approach to drug therapy. They were approved for use because they worked well on the surrogate marker, glycohemoglobin, or blood sugar, even though no studies had been sufficiently lengthy to see whether any of these drugs actually reduced the long-term complication rates. Indeed, no evidence was ever accumulated to show that rosiglitazone successfully prevented major diabetic complications. Ideally, such a drug would not have been used at all for type 2 diabetes, or else it would have been a third-line drug for use in selected recalcitrant cases. Yet with aggressive marketing, Avandia was quite widely used, despite its high cost.
With time, however, worrisome evidence accumulated that rosiglitazone actually increased the risk of serious heart disease. At first the manufacturer simply hid the evidence and attacked independent investigators who attempted to publish such findings (Nissen and Wolski 2007; Moynihan 2010). In the end, the company had to admit the problem and submit to a major warning label. The drug was released in 1999 and by 2006 had reached a sales peak in the USA of $2.5 billion. Once the negative heart disease information was published in 2007, US sales plummeted, but even then Avandia racked up $1.2 billion in worldwide sales in 2009 (Noble 2013). Since the drug would have faced stiff competition from cheaper generic versions after it went off patent in 2012, the company got to enjoy almost a full lifetime of generous profits from a largely useless and actually dangerous drug. Whatever the effect on unfortunate patients, as a business plan, Avandia succeeded splendidly.