Monday, March 15, 2010


The Internet is abuzz with the latest results from a couple of those massive trials that physicians who practice "evidence-based medicine" require before they'll believe in any treatment.

Although I understand why such studies are needed, I hate them, because they're studying a huge, diverse population of patients who may differ a lot in their baseline characteristics, even though the mean is usually all you can see.

Unless the outcome is black and white, for example, 100% of the patients who took the new drug dropped dead within 2 weeks, you need statistics to evaluate the study. Quite often, individual patients may be harmed or helped, but the published conclusion refers only to the average impact, as I noted here. Then physicians apply these average results to everyone.

A good example of this is the blood glucose (BG) arm of the ACCORD study, which was stopped early a couple of years ago because it appeared that the patients who used intensive treatment with a lot of drugs and lowered their A1cs to a mean of 6.5% had higher mortality than those who used standard treatment and had A1cs of about 7.3%.

This was despite the fact that patients in both groups had mortality rates lower than those of most people with diabetes.

In fact, the patients in the ACCORD study were older, had had type 2 for at least 10 years, had other risk factors for heart disease, and started with mean A1cs of 8.3. This means they had probably had poor control for years. Yet doctors are applying the conclusions to everyone.

Many patients are now reporting that their doctors tell them that their excellent A1c levels in the 5s are too low and they should increase them until they're over 7!

Furthermore, like most patients, the ACCORD patients were told to follow an ADA-type diet with less than 30% total fat and less than 10% saturated fat. This means they undoubtedly increased their consumption of carbohydrates, most likely the kind most Americans eat: potatoes, rice, white bread, processed fat-free foods. Yet a recent meta-analysis showed that there is no significant evidence to conclude that saturated fat causes heart disease. Some studies showed an increase when saturated fat was reduced, and others showed an increase. This averaged out to no effect.

The authors suggested that it might depend on what you substitute for the saturated fat, as studies with substitution of unsaturated fat tended to reduce heart disease and mortality and studies with substitution of carbohydrate tended to increase it, although no studies have been done that would actually prove this.

Yet replacing saturated fat with carbohydrate is undoubtedly what people in ACCORD were told to do, and those in the intensive treatment arm of the study got more intensive nutritional counseling and hence probably ate more carbohydrate.

Now the other two arms of the ACCORD study have been published. The blood pressure arm showed that reducing the systolic blood pressure below 120 resulted in no better cardiovascular outcomes than using fewer drugs to keep the systolic blood pressure below 140. The lower blood pressures did result in fewer strokes.

This is the same patient population as the BG arm of the study, and the same caveats apply: longstanding diabetes in an elderly population with coexisting medical problems (34% had already had a cardiovascular event), relatively high starting A1cs and fasting BG levels over 170, and multiple blood pressure drugs given to reach the goal. Also, twice as many of the intensively treated patients gained more than 10 kg during the study.

The final arm of the study was designed to see whether adding a fibrate drug to the treatment of patients already taking a statin would reduce cardiovascular events. The fibrates (they used fenofibrate) reduce triglycerides and increase HDL levels.

Again, they found no significant effect but a suggestion that the drug might help in patients who began with triglyceride levels over 204 and HDL levels under 34. Men appeared to do better and women appeared to do worse on the fibrate. Such studies can show differences that appear to be real but aren't statistically significant.

Again: same population and same caveats.

Another study, the NAVIGATOR study, was reported at the same time. This study started with patients who had prediabetes, with mean A1cs of 5.8 and also either preexisting heart disease or cardiovascular risk factors. They tested whether using valsartan (Diovan), an angiotensin-receptor inhibitor that lowers blood pressure, would reduce progression from prediabetes to diabetes. Similar drugs had been shown in the past to do so.

Again, all the patients were given "lifestyle modification" advice, although the papers don't specify exactly what that was other than the usual ADA line of reducing total and saturated fat and increasing exercise. You have to go to an Appendix, which most people won't read, and then to a reference to a Finnish study they cite to see what type of dietary advice was given.

It turns out to be the usual low fat with "lots of whole grains, fruits and vegetable." Many Americans told to eat lots of whole grains are apt to eat whole-wheat bread (which isn't whole grain) and to drink more orange juice and eat more apples and bananas, and maybe more peas and corn. Very few will up their intake of kale and broccoli and other low-carb veggies.

It turned out that the low-fat high-carb diet plus increased exercise plus the drug reduced the progression to type 2 diabetes from 36.8% to 33.1%, which they calculate is a 13% reduction in the "absolute hazard difference using an exponential model," but a pretty small absolute reduction. It didn't affect the rate of cardiovascular events.

The second arm of the NAVIGATOR trial involved the same patient population and the drug nateglinide (Starlix), which is a sulfonylurea-type drug that increases insulin secretion by the beta cells but for a shorter period than the traditional sulfs.

The rationale was that high postprandial BG levels are said to lead to beta cell deterioration, and higher A1cs are associated with increased heart disease. They tested whether or not this drug would reduce progression from prediabetes to diabetes and whether it would affect cardiovascular events.

They found it did neither.

Do these studies mean there's no point in trying to control our diabetes?

Not at all. What they really show is that you can't give people with longstanding diabetes or even a diabetic tendency and either preexisting heart disease or a lot of heart disease risk factors a low fat, and hence very high carbohydrate, diet, try to control the resulting high BG levels with a lot of drugs, and expect the heart disease to go away.

Furthermore, even though you tell people to eat lots of vegetables and whole grains, you know that in the general population, most of them -- if they modify their diet at all -- will eat high-glycemic foods, low-fat processed convenience foods, and sugary fruits. If they show the dietician that their fat consumption is down, the dietician will probably tell them they're doing great.

No one has tested whether or not trying to control diabetes with lower-carb diets and fewer drugs would reduce heart disease rates.

But I'm afraid that the results of these trials will make a lot of people simply throw up their hands and give up, figuring that heart attacks are inevitable, no matter what they do.

Even if the results from a lower-carb study showed fewer cardiovascular events, I'm afraid most Americans wouldn't make significant changes in their diets. An intelligent woman with type 2 once told me she had trouble eating just a couple of potato chips. I asked why she bought potato chips (she lived alone). She said, "Because I like potato chips."

Well, who doesn't. I also used to like blueberry pie (I probably wouldn't like it now, because it would seem overwhelmingly sweet with relatively little taste) and homemade bread slathered with butter and homemade jam. But I don't eat those things now.

What we need to learn to do is to become gourmets, seeking out foods with a lot of taste and not a lot of carbohydrate, like berries, or exotic fresh vegetables from a farmers market. This is a lot more fun and cheaper than paying $500 a month for a lot of pills to try to cover the damage from eating ho-hum potato chips and packaged snack cakes.

The intelligent people who read this blog will understand this. I worry about the other millions of people in the country who don't have access to good information. I worry about the overworked GPs who don't have time to slog through long statistical studies and try to figure out what an "absolute hazard difference using an exponential model" is.

Many of the details, like the actual dietary advice, in these papers are difficult, if not impossible, to find. If you make an effort to download the full study protocol of the ACCORD study, you find that patients were taught carb counting but it doesn't say how many carbs they were supposed to eat. They were taught self-monitoring of BG, and how to titrate their drugs according to the results. They were apparently not taught how to "titrate" their carb consumption according to the results.

And the authors are often sloppy. For example, sometimes they give both mean and median A1c. Sometimes they give only one. Sometimes they don't indicate which one they calculated.

I worry that the busy physicians will just read the headlines in medical magazines and the New York Times ("Diabetes Heart Treatments May Cause Harm") and conclude that they shouldn't try to treat diabetic patients with high blood pressure, high BG levels, or high lipid levels. Why bother, because they might be sued if they caused harm.

As studies become old, people who write about them tend to simplify, ignoring the many caveats that apply to the studies. For example, Gina Kolata wrote in the recent New York Times
story, "It was discovered 2 years ago that rigorously controlling blood sugar did not prevent heart disease or deaths in people with type 2 diabetes." What that study actually showed was that "rigorously controlling blood sugar with a lot of drugs to cover a high-carb diet did not prevent heart disease or death in elderly patients with preexisting heart disease or at least two cardiovascular risk factors and long-standing poorly controlled diabetes."

But how many physicians have retained Kolata's interpretation? I suspect a lot. I've mentioned the many patients whose doctors told them that their diet-controlled A1cs of 5.6 were too low and they should try to get them up to 7!

I would agree that if someone had an A1c of 5.6 only because they were on 7 different expensive medications with a lot of potential side effects, it would make sense to stop several of the drugs and let the A1c go up a bit, especially if the patient was elderly with several other medical problems treated with even more drugs.

But if someone has an A1c of 4.8 because of strict diet control and a lot of exercise, and if that person doesn't go low (after all, nondiabetics don't go low when they have low A1cs), there's absolutely no reason to tell that person to increase the A1c.

Applying a "rule" for the wrong reasons is the type of faulty logic that has caused harm in a lot of diabetic patients. I know some who have been told by registered dieticians that they should put raisins in their oatmeal "to get the carb counts up."

The reason for the high-carb ADA diet is not to eat a lot of carbohydrate; it's to eat less fat. The idea is that when you eat more carbohydrate, you'll eat less fat. But adding carbohydrate to a meal instead of substituting carbohydrate for fat won't reach the ADA goals (which many people today don't agree with anyway). It will just add calories, increase insulin levels, and promote even more fat gain.

So will patients with type 2 diabetes soon be told to get their blood pressure up, not worry about lipid levels, and pay no attention to postprandial BG levels?

I certainly hope not.

The full texts of the New England Journal of Medicine articles cited are available free here.

1 comment: