Tesofensine, commas, and a not insignificant amount of fat

Today I encountered yet another example of the misleading language I see all too frequently in coverage of science news. I was browsing a health newsletter (the "Pink Sheet") when I saw this:

NeuroSearch pill doubles weight loss, study finds

A Phase II trial of tesofensine found that the drug caused about 10% more weight loss in obese patients compared with placebo and diet. The finding indicates that the treatment, manufactured by Danish firm NeuroSearch, is twice as effective as existing obesity pills, which provide about 5% of weight loss.

Okay, how much weight loss does tesofensine produce - 100% more (double), or 10% more? And more than what in each case? Based on this snippet, I have no idea why this is exciting news, much less what the take-home message is (and a snippet of this size should be ALL take-home message).

So I bopped over to the WaPo , expecting better coverage. This is what I saw in the first couple of paragraphs, which is all most people will probably read:

"Tesofensine produces a weight loss of approximately 10 percent more than placebo and diet in obese patients," said lead researcher Dr. Arne Astrup, from the Department of Human Nutrition, Faculty of Life Sciences, at the University of Copenhagen. Other drugs produce only about 5 percent weight loss, Astrup noted.

Okay: tesofensine doubles the weight loss achieved with other drugs. But if tesofenine only helps you lose a measly "10% more than placebo and diet," why bother prescribing it - much less another drug that gives you even less benefit? Has pharmamania really gotten that bad? It doesn't make much sense, does it?

My hypothesis was that there must be a comma missing in the WaPo article: Tesofensine produces a weight loss of approximately 10 percent [of body weight], more than placebo and diet in obese patients. A 10% reduction in total body weight is a key goal, associated with a variety of significant health benefits, like reduced blood pressure. If other diet drugs typically yield 5% loss of body weight, and placebo yields something less than that, enthusiasm about tesofenine make a lot more sense: it's doubling the effect of other drugs, and more than doubling the outcome you'd get with diet/placebo alone, to help you reach a clinically significant goal.

This explanation is consistent with the details quoted later in the article:

Over 24 weeks, those receiving the lowest dose of tesofensine (0.25 milligrams) lost 14.7 pounds, while those taking 0.5 milligrams lost 25 pounds, and those taking the highest dose (1 milligram) lost 28 pounds. People receiving placebo lost less than 5 pounds.

I'm not great at math, but last I checked, 14.7 is not 10% more than 5! (And neither is 25 or 28!)

But my hypothesis needs testing - on to the original paper, in the Lancet, where to make things more fun for us, the units are British. The abstract says:

The primary outcome was percentage change in bodyweight. . . After 24 weeks, the mean weight loss produced by diet and placebo was 2.0% (SE 0.60). Tesofensine 0.25 mg, 0.5 mg, and 1.0 mg and diet induced a mean weight loss of 4.5% (0.87), 9.2% (0.91), and 10.6% (0.84), respectively, greater than diet and placebo (p<0.0001).

There's my comma! But it's even better than I guessed: it turns out the authors don't simply mean "tesofensine produces 10% weight loss, which is greater than diet/placebo." They're saying that tesofensine produces 10% additional weight loss over and above diet/placebo. That's even more impressive! What the WaPo should have said was not "10% greater than placebo" nor "10%, greater than placebo" but "an additional 10% over and above placebo" - or something like that.

In the intention-to-treat population, change in bodyweight in the placebo group was 2.0% (p<0.0001) from baseline to 6 months, but was greater in all tesofensine groups
(figure 2). The adjusted mean weight reduction above that of placebo was 4.5% (4.5 kg [SE 0.87], p<0.0001) for 0.25 mg tesofensine, 9.2% (9.1 kg [0.91], p<0.0001) for 0.5 mg tesofensine, and 10.6% (10.6 kg [0.84], p<0.0001) for 1.0 mg tesofensine.

Using data from the paper, the placebo group started with an avg weight of 104.3 kg (230 lbs; BMI 35.6) and lost 2.2 kg (4.9 lbs) on placebo, or 2% of body weight. Let's consider the 0.25 mg tesofensine group, who started at 102.7 kg (226 lbs; BMI 35.0) and lost 6.7 kg (14.7 lbs), or about 6.5% of body weight. Their % reduction in total body weight, over and above placebo was 4.5%. Voila. (Just to be explicit, if tesofensine only gave you 10% more weight loss than placebo and diet, as the original newsletter implied, the 0.25 mg group would have lost a mere 2.2% of their bodyweight - about 5.5 lbs, instead of nearly 15!)

Do the same thing for 0.5 mg tesofensine, which is the dosage the researchers recommend, and you get a little more than 9% body weight loss added on top of the 2% achieved by placebo and diet alone. And there's the take-home message.

Whew, that was a lot of numbers. I hope I didn't mix any of them up. But it all goes to show how important clarity is in reporting medical studies. In this case, the poor wording in the WaPo article probably led directly to the inaccurate newsletter article, which totally misrepresented how promising tesofensine is. People generally have trouble with percentages regardless; just ask the typical consumer - or clerk! - to calculate successive clearance discounts in a retail store. That's why when covering health outcomes, you have to be precise about the outcome of the trial, how it was measured, what your units are, and what your basis for comparison is. Saying "Drug A achieves 10% more than Drug B" is just annoying and incomplete.

Incidentally, tesofensine has an interesting story. It's a triple monoamine reuptake inhibitor, preventing presynaptic uptake of norepinephrine, dopamine, and serotonin, which was originally intended to treat the neurological symptoms of Alzheimer's or Parkinson's disease. It wasn't effective enough in those clinical trials to justify continuing into the next phase, but fortuitously, someone noticed unexpected weight loss in some obese patients. A subsequent metastudy indicated that it prompted weight loss without undesirable changes in blood pressure. The new study also showed minimal changes in blood pressure, although there was a slight increase in heart rate in one of the groups.

If tesofensine pans out in further human trials, it'll be another case of a drug working out to treat something rather unexpected - like Viagra - and testimony to the benefits of keeping an open mind in clinical research.

Effect of tesofensine on bodyweight loss, body composition, and quality of life in obese patients: a randomised, double-blind, placebo-controlled trial. Arne Astrup, Sten Madsbad, Leif Breum, Thomas J Jensen, Jens Peter Kroustrup, Thomas Meinert Larsen. The Lancet, Early Online Publication, 23 October 2008. Summary.