Does Caloric Restriction Cause Hormesis: Comments on paper by Turturro, Hass and Hart

Richard Wilson, Ph.D.

Department of Physics, Harvard University, Cambridge, MA 02138

Tel: 617-495-3387, Fax: 617-495-0416


It has been known since man began that starvation leads to underweight and can cause death. It has been known for centuries that overeating (leading to overweight) is bad. Thus any graph of death rate (on the y axis) against weight (on the x axis) will have a minimum at an "optimum weight". While not a symmetric U shape curve, a plot of death rate against the logarithm of weight (on the x axis) is more symmetric. This simple, old and obvious observation is often forgotten.

I never quite understand what people (or why) are talking about when they use the word hormesis. Nor am I appreciably helped by the first of the two definitions of Turturro et al. So I will start with his second, broader definition and discuss how the data themselves might narrow it. Death (which is clearly an adverse response) has a non monatonic relationship with weight and weight is clearly therefore, according to this definition, hormetic.

Several questions arise when we try to understand this issue further.

(1) Is the same curve (of death rate vs. weight) applicable for each cause of death or is it a superposition of a two (or more) curves: rate of one cause of death falling with body weight superimposed on death by another cause increasing with weight?

(2) What is the direction of causality? Is low body weight the cause of death or is a lingering disease such as cancer (a harbinger of death) the cause of the decreased body weight?

(3) Is body weight itself the best descriptor or is it a surrogate for another cause such as food intake? and if food intake is it calories, fat, or accompanying pollutants?

(4) Is there a changed meaning when we change the sign of the variable on the x axis and discuss weight loss instead of weight gain?

Let us address these questions in turn. The Framingham (1959) study (among others) clearly showed that heart problems of all sorts increase with weight (above an optimum value), and less clearly that cancer rates increase also. It is usually assumed that causality proceeds from weight increase to heart disease or cancer. Less clearly there is an association between a low body weight and cancer; in this case it is usually assumed that the cancer has led to a wasting away of the body and reduced weight - causality proceeds in the opposite direction.

These early observations in people have been well confirmed in animal studies. Until recently rodent bioassays were usually carried out by feeding the animals ad libitum. But numerous tests, and in particular the extensive set of tests carried out at the National Center for Toxicological Research (NCTR) described by the authors and others, showed that when feeding was limited, the weight is less, the animals lived longer and get fewer tumors. Indeed the effect is so striking and so pervasive that many analysts argue that the animal bioassays (in particular those of the National Toxicology Program) should always be done with a carefully controlled diet or else the bioassays are less easily interpretable. The effect is to add an unwanted fluctuation, "noise" in electrical engineering parlance, to the data and therefore will show spurious carcinogenic (and anticarcinogenic) responses if the usual criteria for statistical significance are applied.

But is the weight the important parameter or is it a surrogate for the real cause? Or mathematically speaking what is the proper independent variable and what are the dependent ones? This is less easy to tell. It is conventional to associate weight with food intake - although exercise probably reduces weight for the same food intake. Health specialists (without a justification that I understand) keep telling us that exercise is more important for longevity and health generally than reducing food intake but that would not be true if food intake were the appropriate independent variable. As noted by Torturro the actual data on this subject are confusing. The effect of exercise should be capable of being tested in a well controlled animal study but that does not seem to have been done. But even the assumption that weight is only a surrogate for food intake, does not tell us what aspect of food intake is important: fat; calories or toxic pollutants present in roughly equal amounts in all foodstuffs. Each of these has been proposed with some logical justification at one time or another. It is here that the work at NCTR is particularly important. Scientists at NCTR argue that it is the calories ingested that are important and NOT the fat, and NOT the toxic pollutants. This is a conclusion of tremendous importance. It changes completely the naive picture (that the whole problem is the wicked capitalist industry) of who and what is most responsible for cancer.

If we accept (as I believe we must) the conclusion that excessive caloric intake is the independent variable responsible for an increase in cancer and too little caloric intake is responsible for immediate death we clearly have caloric intake as a hormetic agent according to Torturro's second definition. What happens as we change the sign of the independent variable and talk about caloric restriction? We could say that since there is a U shaped curve between death and the caloric intake changing the sign of the intake will still leave us with a U shaped curve and therefore hormesis. But here I suspect that it is useful to recognize that the relationship is a better U shape if the logarithm of the intake is taken. This is also true if the sign is reversed. This suggests that (1/intake) is a better variable than (- intake). This is not entirely semantics. When we talk about dietary restriction we must state restriction from what? Presumably a reduction from some (undefined) optimum dietary intake. We are back to the early discussion of the Framingham study. What is the optimum weight for a person of a given size? Most people would say that it is a little below the minimum of the death vs weight curve athough most individuals would only accept a couple of pounds below one's actual weight. These considerations make careful definitions very important. Torturro et al. make this even more complex limiting their phrase "caloric restriction" to mean less than a restriction giving outright starvation. But they fail to give a reliable recipe for determining what that point is.

But when I return to an attempt to understand useful distinctions between different definitions of hormesis I am left with the question: does caloric intake cause some adverse effect on some biological end point at high doses IN THE SAME WAY (i.e. by the same biological mechanism) and with the same direction of causality? I find that the discussions of this are obscure and am not yet convinced that hormesis in this sense occurs whether by caloric intake, restriction or indeed any other independent agent.


Build and Blood Pressure Study (1959) Society of Actuaries, 208 S. LaSalle Street, Chicago 4, IL

Build Study (1979) Society of Actuaries and Association of Life Insurance Medical Directors of America, 208 S. le Seet, Chicago 4, IL