Pinel JP, Assanand S, Lehman DR.
Hunger, eating, and ill health.
Am Psychol. 2000 Oct;55(10):1105-16.
Three studies of human subjects living in wealthy countries suggest that low levels of consumption can have beneficial effects on health. Two of the studies, the Okinawa study (Kagawa, 1978) and the Nurses' Health study (Manson et al., 1995), were correlational studies demonstrating that health and longevity are greatest in individuals whose levels of consumption and body weights, respectively, are below prescribed national norms. The third study, the Biosphere 2 study (Walford, Harris, & Gunion, 1992), was an uncontrolled experimental demonstration that an improvement in health can result from a substantial reduction of food intake in ostensibly normal-weight, healthy individuals.
In the Okinawa study (Kagawa, 1978), the caloric intake and health of inhabitants of the Japanese island of Okinawa were compared with those of inhabitants of other parts of Japan. Adult Okinawans were found to consume, on average, 20% fewer calories than other adult Japanese, and Okinawan school children were found to consume 38% fewer calories than recommended by Japanese health authorities. Remarkably, the rates of morbidity and mortality in Okinawa were found to be markedly lower than in other parts of Japan, a country in which overall levels of caloric intake and obesity are well below Western norms (see Kagawa, 1978). Most notably, the death rates from stroke, cancer, and heart disease in Okinawa were only 59%, 69%, and 59%, respectively, of those in the rest of Japan, and the death rate for Okinawans between 60 and 64 years of age was only 50% of the Japanese average. Indeed, the proportion of Okinawans living to be over 100 years of age was 5 to 40 times greater than that of inhabitants of various other areas of Japan.
In the Nurses' Health study (Manson et al., 1995), the association between body-mass index and mortality was assessed in a large cohort of U.S. female nurses-levels of consumption were not directly measured. In this study, over 100,000 female nurses were recruited and studied over a 16-year period. In contrast to previous studies of the relation between body-mass index and mortality (see, e.g., Stevens et al., 1992; Tuomilehto et al., 1987; Wilcosky, Hyde, Anderson, Bangdiwala, & Duncan, 1990), this study controlled the confounding effects of cigarette smoking, which is more prevalent among relatively lean people (Garrison, Feinleib, Castelli, & McNamara, 1983). In addition, it controlled the effects of several other potential confounds, including age, levels of alcohol consumption, levels of dietary fat intake, menopausal status, preexisting disease, illness-related weight loss, and levels of physical activity. Most notably, the results revealed a positive correlation between body-mass index and mortality, with the lowest mortality rate occurring among those nurses with body-mass indices below 19-that is, among those nurses weighing at least 15% below the average weight of U.S. women of a similar age and at least 10% below their recommended weights according to the widely used Metropolitan Life Insurance Company Table of 1983. Furthermore, negative correlations were observed between body-mass index and various measures of health: Diabetes, gall stones, hypertension, and nonfatal myocardial infarction were all less frequent in the leanest nurses than in the normal-weight or overweight nurses. Apparently, the various health advantages of a low body-mass index had not been detected in previous studies (e.g., Tuomilehto et al., 1987) because they had not controlled for cigarette smoking.
In the Biosphere 2 study (Walford et al., 1992), four normal-weight females, three normal-weight males, and one mildly overweight male lived continuously for two years in a self-contained ecosystem: the so-called Biosphere 2. During their first six months in Biosphere 2, the eight participants were limited to a diet of 1,780 calories per day, an intake nearly 600 calories less than the current North American average (Walford & Walford, 1994). Among the health benefits observed in participants following this six-month period were decreases of 15% in average body weight, 30%/27% in average blood pressure, 38% in average cholesterol level, 20% in average fasting blood glucose level, and 24% in average white blood cell count.
The results of the aforementioned studies have been the focus of attention (e.g., Weindruch, 1996) because of their provocative implication that consumption below levels prevalent in wealthy countries can improve health and longevity. However, because they were uncontrolled, these three studies, even when considered together, provide only equivocal support for this conclusion. For example, it is impossible to rule out the possibility that the effects on health in each of the three studies were mediated by differences in the constitution of the high-calorie and low-calorie diets, rather than by total caloric intake per se; because the constituents of the high-calorie and low-calorie diets were not equated in any of the three studies, the health gains may have resulted from the fact that the lower calorie diets were richer in health-promoting nutrients and lower in toxins and fats than the high-calorie diets. However, this and other methodological problems (e.g., self-selection) have been resolved by controlled experiments on dietary restriction in nonhuman species.