Front-page headlines last month said that women who drank alcohol gain less weight. This would be true if, except for their drinking, nondrinkers were identical to drinkers at the beginning of this 13-year Archives of Internal Medicine study.

But Timothy Naimi of the Centers for Disease Control and colleagues looked at 30 confounding factors in 250,496 adults that could adversely affect future health and weight, such as diabetes, no health insurance, divorce, poor education, low income, poor health and no exercise, and found 27 to be more common in nondrinkers. So if nondrinkers from the outset are more sedentary, or more stressed from divorce or diabetes, you can bet that over the next 13 years they'd gain more weight. Just focus on alcohol, and you'd mistakenly conclude that alcohol caused less weight gain.

Remember when estrogens reduced heart disease? They improve cholesterol, and in 1985, the Nurses Health Study of 32,000 women found that those on hormones had 50 percent less heart disease.

But a later study showed estrogens increase heart disease in older women. Why do we believe this one? Because it was a much more accurate randomized, experimental study, while the 1985 one was an observational one that followed two groups of women — those who decided to take estrogen and those who didn't.

Those on estrogen exercised more, ate better diets and got better medical care than those who did not. These confounding factors were the real reasons they had fewer heart attacks, and likely why drinkers gain less weight.

In randomized studies, folks are randomly assigned to two groups: one takes the real drug, drink or herb and the other takes a fake one (placebo). You then compare the two groups after days to years. Only randomized studies can eliminate confounders and show the true effects of our potions. The Archives study finding drinkers gain less weight is an observational study. As is every study claiming that alcohol reduces heart disease or prolongs life.

Indeed, the Framingham Heart Study found that folks who drank more alcohol gained more weight over the next 20 years. With observational studies, you can never be sure you've properly accounted for all the confounders. The Archives study tried to correct for confounding factors, but the 1985 hormone study mistakenly found 50 percent less heart disease even after correcting for obesity, smoking, cholesterol and diabetes. Alcohol has twice the calories of rice, so if I had to guess what drinking wine would do in a randomized study, I predict it would stimulate the palate and make women fatter.