Endodoc

This week’s New England Journal of Medicine (Volume 359, No. 3, July 17,2008) contained an article entitled “Weight loss with low-carbohydrate, Mediterranean, or low-fat diet,” written by Iris Shai and colleagues. Although the study was conducted in Israel, it was a large international effort with co-investigators from all around the globe. The study results even made the New York Times (“Healthy diets shown to have benefit despite modest weight losses,” written by Tara Parker-Pope, Thursday, July 17, 2008, page C3).

The Study Design

The purpose of the study was to evaluate the effectiveness of several different diet plans long-term with an effort to minimize “drop-outs.” The investigators claimed (correctly) that most previous studies had been hampered by their short-term nature and/or large drop-out rates. The plan was to recruit slightly obese subjects from a single work-site and monitor them for 2 years. The investigators recruited 322 subjects willing to be randomly assigned to 1 of 3 dietary regimens: a low-fat, restricted calorie plan (group 1) ; a Mediterranean, restricted calorie plan (group 2) ; and a low-carbohydrate, non-restricted calorie plan, based on the Adkins diet (group 3). The mean age of study subjects was 52 years and the mean BMI was 31 (I’m sure you remember from previous entries that the BMI is calculated by dividing the weight in kilograms by the height in meters X height in meters; normal is <24.9, overweight is 25-29.9, and obese is 30 or more). Eight-six percent of subjects were male. The study subjects had intensive education regarding the diet plans and frequent follow-up.

The Results

First, adherence to the regimens (I think that meant that the study subjects showed up for their follow-up visits?) was pretty good, about 95% after 1 year and 84.6% after 2 years (272 out of 322 subjects). For each study group the maximum weigh loss was between 1-6 months into the study. At the end of the study, the mean weight loss in subjects who completed the study was 3.3 kg, 4.6 kg, and 5.5 kg in groups 1-3, respectively. The mean BMI fell by 1 in group 1 and 1.5 in groups 2 and 3. There was a difference in results comparing males and females; the 45 females in the study lost 0.1 kg in group 1, 6.2 kg in group 2, and 2.4 kg in group 3.

The investigators found no particular safety problems with any of the diet plans. The subjects with diabetes mellitus fared best with the Mediterranean diet plan in terms of improved fating plasma glucose values (the improvements in hemoglobin A1cs were about the same in the 3 groups, about 0.5%). Lipid profiles showed modest improvements with all 3 diet plans but group 3 (the low-fat diet group) generally showed the greatest improvements.

The investigators concluded that based on their results, Mediterranean and low-carbohydrate diets may be effective alternatives to “classic” low fat diets. They also concluded that a workplace medical care model for weight reduction such as in their study might be a useful way to achieve beneficial results and that even modest weight reductions can improve risk factors for undesirable health outcomes. The investigators did not really focus on the benefits of one diet plan vs. another (I do wonder what the first draft of the paper submitted to the journal looked like, but we’ll never know).

What do the results mean?

I found it somewhat difficult to interpret the results. First, the investigators should have presented the male and female data separately given the fact that 86% of the subjects were male but also given the results in the 45 female participants; the females lost quite a bit more weight on the Mediterranean diet (average of 6.6 kg) than on either of the other 2 plans. In fact, the females lost only a trivial amount on the other 2 plans. I have no idea what the data mean except that we should think twice before trying to generalize the study results to males and females alike.

The investigators are to be commended for having a pretty low drop-out rate but even 15% is of concern. In addition, the data showed that the drop-out rate differed among the 3 groups- it was highest in the low-carbohydrate group (22%) and lowest in the low-fat group (9.6%) with the differences statistically significant.

Finally, even two years of follow-up is far too brief a period of time to come to any firm conclusions about long-term effects of the of the 3 diet plans with respect to efficacy and safety.

What do the results really mean?

If one takes a hypothetical study subject with a BMI of 31, it might be a male about 175 cm tall (almost 69 inches tall) who weighs 95 kg (209 lbs). To attain a weight at the upper limit of the “normal” BMI range (24.9) he would need to lose about 18 kg, far more than was achieved in the study conducted by Shai and colleagues. Clearly, even heroic efforts to achieve weight loss by dietary manipulation as undertaken by the investigators still fall far short of what is desirable. But, still, every little bit helps. I do wonder what the results would have been in study subjects with much higher BMIs?

In my last entry I provided some tips for non-experts on how to make sense out of medical journal articles. The main purpose of the entry was to provide a framework for analyzing the results of 2 important diabetes studies that were published in the New England Journal of Medicine about 3 weeks ago (Volume 358). Accompanying the reports were 3 editorials, so you can tell the studies were either very important or very controversial.

The first study was called ACCORD, which stands for Action to Control Cardiovascular Risk in Diabetes; the second was called ADVANCE, which stands for Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation (I don’t know how clinical trials come up with these bizarre titles- maybe they have a contest and give a prize for the title that provides the best acronym?). Both studies were large-scale clinical trials designed to assess the effects of “tight” glycemic control on cardiovascular outcomes in people with type 2 diabetes mellitus.

The Study Designs

In many ways the 2 studies were very similar. Both studies recruited large numbers of patients with type 2 diabetes (10,251 people in ACCORD and 11,140 in ADVANCE). The average ages and durations of diabetes were fairly similar in the 2 studies (average age was about 60 years and average duration of diabetes was about 10 years) and both had quite a few more male than female participants (about 60:40 ratio). In both studies, patients were randomly assigned to either intensive or standard therapies. The goals of the intensive therapy was to bring hemoglobin A1c levels as close to normal as possible- less than 6% in ACCORD and 6.5% or less in ADVANCE (remember, that’s the test that more-or-less averages the blood glucose level over the previous 3 months or so). In both studies, the primary research question was whether intensive therapy (that was designed to achieve blood glucose levels as normal as possible) would decrease cardiovascular outcomes compared to standard therapy. Patients were eligible for the studies only if they had prior cardiovascular events or had cardiovascular risk factors (in addition, patients were eligible for the ADVANCE study if they had only prior major microvascular events).

The treatment strategies differed in the 2 studies. In the ADVANCE study, patients in the intensive treatment group all received gliclazide, a sulfonylurea drug and any other medications selected by the treating physicians but no sulfonylureas besides gliclazide (a major sponsor of the study was the pharmaceutical company that makes gliclazide). Patients in the standard group could be treated with whatever medications the treating physicians selected but no gliclazide. In the ACCORD study, patients were treated with whatever medications the treating physicians selected but only certain medications were provided free-of-charge by the study (the drug list did not include any insulin preparations).

Study Results

The ACCORD study was stopped after 3.5 years when results showed a statistically increased risk of death in the intensive treatment group with 257 vs. 203 deaths in the standard treatment group. Average hemoglobin A1c in the intensive group was 6.4% vs. 7.5% in the standard group- the average was about 8.1% in both groups at study onset. In terms of well documented cardiovascular outcomes, the intensive group actually had fewer than the standard group but the difference was not statistically significant. Of course, it is likely that the excess deaths were actually cardiovascular in nature. It is interesting that patients in the intensive treatment group with hemoglobin A1cs less than 8% who had no major cardiovascular events prior to entering the study before entering the study, actually had a significantly lower risk of fatal or nonfatal myocardial infarction than patients in the standard group.

In the ADVANCE study, median follow-up was 5 years. Average hemoglobin A1c in the intensive treatment group was 6.5% vs. 7.3% in the standard group. Patients in the intensive group showed a statistically significant decrease in the risk of developing a composite of major macrovascular and a microvascular outcomes, primarily related to a decrease in the risk of developing kidney disease in the intensive therapy group. There was no statistically significant decrease in the risk of developing major macrovascular disease, cardiovascular death, or death from any cause in the intensive treatment group compared to the standard treatment group.

Now What?

So, here we have 2 large-scale clinical trials in patients with type 2 diabetes with somewhat different results: In the ACCORD study, intensive treatment patients were more likely to die than standard treatment patients, while in the ADVANCE study, patients in the intensive group were less likely to develop kidney disease but not more or less likely to develop cardiovascular disease or die than standard treatment patients. Whom are we to believe?

First, were the increased deaths in the ACCORD study intensive group patients “real” or just a statistical fluke- if the study had gone longer or if it had included more patients, would the results have been the same? Based on the statistical analysis, there was about a 1 in 20 chance the results were statistically different between the treatment groups only by chance. We’ll probably never know.

Second, if the increased deaths were “real,” how come? Were there any differences in the design of the two studies that might explain the differences in results? In my opinion, there were major differences in the design of the 2 studies. Most important, the diabetes medication profiles between the 2 studies and even between groups in the ACCORD study were strikingly different. The use of thiazolidinediones, drugs already implicated in increased risks of cardiovascular deaths and deaths from any cause, were used heavily in the ACCORD study but not in the ADVANCE study. In fact, those drugs were used much more in the ACCORD intensive treatment patients than in the standard treatment patients. That fact alone raises serious questions about the meaning of the ACCORD results- were the increased deaths in the ACCORD intensive group patients somehow the result of the lower average blood glucose levels or the result of the medications employed to achieve those blood glucose levels?

Should We Have Expected To See A Benefit From Intensive Therapy?

I hate to further complicate things, but sitting in an armchair I would not have expected to see much of any differences between the treatment groups in either of the studies. In a number of earlier entries I have discussed the results of the landmark Diabetes Control and Complications Trial- the study that first showed in convincing fashion that intensive therapy prevented diabetes complications. In that study, which included only patients with type 1 diabetes, the intensive treatment group maintained hemoglobin A1c at about 7.3% with the standard group at about 9%. It took 9 years to show meaningful differences in outcomes between the 2 groups. Further, the data showed that risk for diabetes complications was directly related to hemoglobin A1c levels. The relationship was not, however, linear; the relationship was exponential- in the hemoglobin A1c range 6-8% there was a slowly increasing risk for complications as the hemoglobin A1c increased which rose rapidly beyond 8% or so. Thus, lowering the level of hemoglobin A1c from 7.5 to 6.4% as in the ACCORD study and from 7.3 to 6.5% as in the ADVANCE and for only a few years would not have been likely to show much of a difference in results between the treatment groups (in the DCCT, it took 24 years to show a clinically and statistically significant difference in cardiovascular outcomes between the treatment groups!).  There was a statistically significant decrease in the risk of developing diabetic kidney disease in the ADVANCE intensive treatment group compared to the standard treatment group but the benefit was tiny (18 vs. 20% risk).

Where Do We Go From Here?

In my opinion, these studies taken together illustrate the dangers in making patient care decisions based on the results of clinical trials. You should have seen the headlines- “study results turn medical thinking on its head,” with “experts” telling us that perhaps “tight” glycemic control in diabetes is not as good as “semi-tight.” In fact, we should be very cautious about coming to any conclusions based on the ACCORD and ADVANCE studies despite the investigators’ best intentions. Don’t say I didn’t warn you in my last entry about the pitfalls of believing what you read.