Is low-fat milk good for you and your family? Yet another study says no.

The start of the low-fat craze back in the 1980’s perfectly matches up with the start of the obesity epidemic. A coincidence? Probably not.

Low-fat products usually contain more sugar and more starches. If not you’ll probably end up eating more carbs anyway as you’ll be hungrier. This raises the levels of the fat-storing hormone insulin. Study after study prove that low-fat diets are worse for our weight than high-fat low-carb diets. The same is true for kids.

Not surprisingly, a recent Swedish study showed that people using low-fat dairy products end up gaining *more* weight. Now a new American study shows the same thing. Kids drinking low-fat milk are not only more often obese, they also tend to keep gaining more weight than kids drinking full-fat milk:

LA Times: **Low-fat milk doesn’t help toddlers’ weight, study says**

When will the disastrous low-fat craze end? How many more kids are going to get obese for no good reason?

What do you think?

"to our advantage"

Absolutely. Humans have evolved the capacity to drink milk at least 4 times independently in various populations. It's clearly a tremendous nutrition advantage for those of us with those genetics. And I don't intend to throw it away. If you have the genes to tolerate milk why not? To argue that "we're the only species that does it," as if that's a problem, is just dumb. We're the only species that uses an alphabet too - so do we have to stop? It's just a ridiculous statement prima facie.

So here are my comments: first, Ted Hutchinson (comment number 6) is perfectly right about somatotropin hormone use. But even when you take this into consideration, it is impossible to back any of the conclusions of the authors. What the lay press stated proves they know nothing about biostatistics and epidemiology.

Now, an exerpt of the results section of the abstract (to keep this as simple as possible). Sorry, it is dry as can be.

"Compared with the reference (0 to <0.5 servings/day), those consuming larger amounts of high-fat dairy had higher breast cancer mortality (0.5 to <1.0 servings/day: hazard ratio [HR] = 1.20, 95% confidence interval [CI] = 0.82 to 1.77; and ≥1.0 servings/day: HR = 1.49, 95% CI = 1.00 to 2.24, P trend = .05), higher all-cause mortality (P trend < .001), and higher non–breast cancer mortality (P trend = .007); the relationship with breast cancer recurrence was positive but not statistically significant. The higher risk appeared consistent across different types of high-fat dairy products.

Conclusions: Intake of high-fat dairy, but not low-fat dairy, was related to a higher risk of mortality after breast cancer diagnosis."

What does this gibberish mean?

First, let's look at the Confidence Interval (CI). If the confidence interval contains "one", this means that there is no difference between what is being compared.

So: Higher breast cancer mortality? Let's see:

For 1/2 to 1 serving per day, their number is 1.2 (theoretically 20% more than the reference) UNTIL you look at the confidence interval: .82 to 1.77 - this contains one, which means there is strictly no difference between the groups being compared.

For one or more serving per day, the calculated risk is 1.49 (49% more!), once again UNTIL you look at the confidence interval which is 1.00 to 2.24. once again, it contains one, which means there is no difference between what is being compared.

In the abstract, which is what most physicians and dietitians will read because they are unable to interpret the biostatistical part of the article, the authors cut corners and simply state that "consuming high fat dairy will bring "higher all-cause mortality (P trend < .001), and higher non–breast cancer mortality (P trend = .007)". No explanation.

So instead on relying on these numbers, I went to the article to look at results. What do they show?

Higher all cause mortality: for the 1/2 to 1 portion per day, 1.05 - 5% more risk - (CI .88 to 1.53) - contains ONE, therefore not a significant finding. For the more than one portion per day, 1.64 (CI 1.24 to 2.17, the first significant result in this study.

Higher non breat-cancer mortality? 1/2 to 1 portion per day: 1.06 (CI .71 to 1.59, not significant) and for the more than 1 portion per day, 1.67 (Confidence Interval 1.13 to 2.47). This is also statistically significant. So what do these two positive results mean? Not very much. In this study where most results are not statistically significant, it is tough to make a strong statement. Those positive results could be caused by something else than fat, the added hormones hinted to by Ted Hutchinson for example.

Now: it is important to understand the p value. I adapted here an explanation from askville.amazon.com: their texts on the subject were quite nice and undestandable, so I decided to modify and expand on their text rather than re-writing an explanation from my notes.

First, an important fact that is unknown to virtually everyone: with biostatistics, it is impossible to prove anything. Curiously, we can nevertheless “prove” that things are false. So since we cannot prove that things are true, we try to prove that the contrary of what we want to prove is false. That is what biostatistitians call the "null hypothesis".

A p-value is the statistical evidence against this null hypothesis. It does not tell you that the null hypothesis is true or not, it only tells you that there is significant evidence to reject it or not (because those results could be due to mere chance). Commonly a p-value under 0.05 is considered significant. With a p-value of 0.05, that means that there is only a 5% risk that the values that have been found could be due to chance. 5% risk of findings caused by chance is considered acceptable. Over that, it is much less acceptable. The true p-value is then calculated according to a number of factors, including the type of statistical distribution. If the calculated p-value is greater than the arbitrarily chosen value, then we say that the null hypothesis has to be accepted (we have to accept as true the contrary of what we wanted to prove). It is only when the calculated p-value is at 0.05 or lower that we reject the null hypothesis and therefore accept our hypothesis… Are you still following me?

In the study that “proves” that high fat milk products increase the odds of recurring cancer death and other horrible things, here are some calculated p-values: for recurrence of breast cancer, .18 (18% odds that the findings could be caused by chance). For breast cancer deaths, p-value is .82 (82% odds that this finding could be caused by chance), death by all causes, p-value .94 (age adjusted) (94% odds that the finding could be caused by chance). These findings must be rejected. They mean nothing because the odds of them being caused by chance are much too high.

Got it? We have numerous confidence intervals in this study that mean that there is NO DIFFERENCE between the reference group (no high fat) and the fat consumption we are looking at. We also have numerous p-values that mean that the findings have to be rejected because the odds of these findings could be caused by chance alone are too high.

Biostatistics and epidemiology are not exact sciences. Nutritional research is very difficult for the patients (try to fill out one of those questionnaires) and for the researchers who have to come up with a meaningful analysis of the data, while taking into account a number of other factors that could be the cause of what is observed. Researchers also do their analysis with their own biases (we all have some). So before believing anything that is written in the lay press, please go to the original article or read critiques done by people who can interpret this gibberish (www.dietdoctor.com is a very good place to go for this kind of critique!) But I guess I'm preaching to the converted...