| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||
1 Laboratory of Experimental Gerontology
2 Research Resources Branch, Intramural Research Program, Gerontology Research Center, National Institute on Aging, National Institutes of Health, Baltimore, Maryland.
Address correspondence to Donald K. Ingram, PhD, Laboratory of Experimental Gerontology, NIA, NIH, 5600 Nathan Shock Dr., Baltimore, MD 21224. E-mail: ingramd{at}grc.nia.nih.gov
DIET restriction (DR) remains the only reproducible, nongenetic intervention for alteration of aging processes including extension of average and maximal life span, prevention and/or amelioration of many age-related diseases, and maintenance of more youthful physiological function later into life. A substantial literature has made it clear that these effects of DR are observed in nearly all species in which it has been tested (1,2). Despite this impressive body of experimental results, it remains unknown whether DR can impact positively upon human aging. Indeed, whether DR is relevant to human aging remains one of the most significant unanswered questions in contemporary biogerontology.
Nonhuman primates have long been used as models for human disease including aging (3,4). To this end, the recent article by Bodkin and colleagues (5) reporting on age-related morbidity and mortality in rhesus monkeys and the putative beneficial effects of DR provides an interesting, though preliminary, assessment of the possible relevance of reduced calorie intake to human aging.
In general, Bodkin and colleagues concluded that DR increases average age at death associated with prevention of hyperinsulinemia and the mitigation of other age-related diseases and disease markers. Caution must be exercised to avoid over-interpretation of these findings. We contend that their results do not provide sufficient evidence to support the conclusion that DR exerts the same widespread beneficial effects in nonhuman primates (and by genetic similarity, human primates) as has been extensively reported in rodents over the last 75 years. The preliminary nature of these findings is dictated by methodological issues, including study design, subject characteristics, and statistical methodology, any of which may limit the generalizability of these findings.
Of significant concern is that several experimental controls typically used in rodent DR studies were not applied in the Bodkin and colleagues study. First, only 8 of the total 117 rhesus monkeys were in the DR group, and only 3 deaths occurred in this experimental group. This is an extremely small sample size for any mortality study. The typical rodent longevity study would include 30–40 animals per treatment group. Second, while the number of monkeys in the ad libitum (control) group is sufficiently large (n = 109), control and DR monkeys did not enter the study at the same time, nor were they randomly assigned to either the control or DR groups.
It is apparent that the study by Bodkin and colleagues was not designed at the outset to address the question to which the conclusions are directed, specifically the effects of DR on aging and age-related disease. The authors state that all subjects were part of a longitudinal study on aging, obesity, and spontaneous diabetes. Monkeys described in this paper were part of a series of individual experiments (6–15) rather than a single focused investigation of DR and aging. The 8 DR monkeys were brought into the larger colony as a separate cohort of adult monkeys already on a weight stabilization protocol designed to maintain a normal adult body weight. For the most rigorous test of the hypothesis that DR retarded aging, environmental and procedural variables would also need to be controlled. Specifically, all monkeys would need to be housed in the same type cage, in the same or similar vivaria, and decisions regarding treatment of disease and euthanasia would have to be standardized. This type of environmental control was not described by Bodkin and colleagues. Thus, rather than a rigorous experimental design addressing the question of the effects of DR on mortality and morbidity, the current report would be more accurately described as a meta-analysis (though lacking appropriate statistical methodology) of many studies over several years from the same laboratory.
Another methodological issue is that the paper does not make clear how many animals had a known date of birth and therefore a reliable age at death. The paper states explicitly that all animals had a recorded age (p. 213). The practice of recording an estimated age when the specific birth date is not known, particularly for wild-caught monkeys, is common in primate research. Therefore, monkeys frequently have a "recorded" age, but not a known date of birth. Over the 24-year span of this study, methodology for age estimation has no doubt improved, and it is likely that several individuals were responsible for these age estimations. Certainly these factors contribute to significant variability in estimated ages thereby negatively impacting any conclusions about differences in life span and age at onset of disease.
Differences in the diet fed to monkeys in the study could have led to another methodological problem. Unlike most rodent DR studies in which diets for control and DR groups are highly controlled, it is apparent that monkeys in the Bodkin and colleagues study did not all consume the same diet. Standard monkey chow (17% protein, 70% carbohydrate, and 13% fat) was fed to most animals in the study. However, 15% (n = 16, p. 213) of the AL monkeys were fed a liquid diet different in composition (14% protein, 55% carbohydrate, and 31% fat) (8) and most notably over twice the amount of fat compared to the standard diet.
Certain characteristics of subjects included in this experiment are also cause for concern. The authors point out that body weight (10–11 kg) and percent fat (17%–24%) for the DR group was comparable to that of a normal, lean monkey (p. 213). For DR monkeys the experimental intervention was to maintain that weight throughout the study. The reader does not know what amount of food was provided to each monkey in this weight stabilization paradigm, and no information on calorie intake is given for AL controls. Body weights on entry into the study were comparable in the AL group (Table 1, p. 214); however, data on weight or fat percentage at the end of the study is not provided. It is unclear whether DR monkeys gained or lost weight during the study, and, more importantly, the extent of weight gain and increased adiposity in AL controls is unknown. Previous publications from this laboratory suggest a body weight range of 15.4 to 19.5 kg, and fat percentages ranging from 20% to 48% (16) would be expected in the AL-fed monkeys that were featured in this report. Thus, it is possible that a significant proportion of the monkeys in the AL group were morbidly obese at death or at time of disease onset. This concern calls to question whether monkeys in the AL group represented a relevant group for comparison to the 8 monkeys in the DR cohort.
An additional concern is the insulin levels for subjects at entry into the study (Table 1). At entry, "normal" controls had insulin levels that were 3 times higher than DR monkeys. Monkeys in the hyperinsulinemic and diabetic groups had insulin levels that were 7.5 and 4.5 times higher than DR, respectively. Given these differences among subject groups, it is possible, and indeed likely, that the differences between AL and DR groups in this study were related more to excessive weight gain and adiposity in the AL group. It is therefore difficult to conclude that the morbidity and mortality findings were due to beneficial effects of DR in the 8 monkeys on weight stabilization. A more appropriate conclusion (or at least deserving of mention in the paper) would focus on the harmful effects of obesity and hyperinsulinemia in the control group, which has been the main focus of research in this group.
Beyond these methodological concerns, the statistical analysis of the mortality data should also be interpreted with caution in the Bodkin and colleagues study. The authors used a Cox proportional hazards regression model to assess differences in mortality between the DR and AL-fed groups. Table 1 indicates that the animals' age of entry into the study is quite varied, ranging from 4 years in the AL normal group to 29 years in the AL diabetic group. In almost all studies of survival with mortality as the outcome, age is an important covariate strongly related to mortality and, in many cases, age is related to the nonproportionality of many other risk factors. Thus, ignoring age and accepting the proportionality assumption can lead to conflicting and nonconclusive results. It is also puzzling why the authors report different analyses using different reference groups (Table 2). Surely some justification and explanation for doing this should have been given in the paper. The 95% confidence intervals for the hazard ratios presented in Table 2 show that there are essentially no statistical differences in these ratios from the no-effect ratio reference values of 1.0. Also, many of the confidence intervals are quite large, indicating the uncertainty of the results. The only significant result was the DR group versus the AL hyperinsulinemic group, which had the widest confidence interval of all those reported in Table 2. Most importantly, using the group with the largest sample size, AL normal, as the reference group resulted in no evidence of statistical significance at all. In the abstract of the paper, the authors state the following: "Compared with the DR monkeys, the AL monkeys had a 2.6-fold increased risk of death." Although, in the body of the text, the authors acknowledge that this comparison was not statistically significant, they continue to make similar statements throughout the paper. If anything, Table 2 serves to demonstrate the uncertainty of the results and reflects the limited amount of data presented in the paper, especially for the DR group.
Further concerns not addressed by Bodkin and colleagues involve the issue of censoring monkeys from statistical analyses. Are the censoring patterns the same in the different study groups? If not, then any informative censoring might be related to the outcome. Finally, Table 3 shows that there are no differences in the risk of mortality between the DR and AL groups when other characteristics or covariates are entered into the regression model. Thus, the statistical results presented by Bodkin and colleagues do not appear to support any finding of differences in the risk of mortality between the different dietary groups. If anything, the authors' results reflect the unplanned or convenient nature of the study population characterized by the small sample size in the DR group and the overall heterogeneity of the animals in the study.
In conclusion, the report by Bodkin and colleagues yielded interesting data on the relationship of nutrition to aging and age-related disease. However, certain conclusions reached in this report must be interpreted with caution. The methodological and statistical concerns outlined above make it problematic to conclude that DR in longer-lived species, such as rhesus monkeys, will exert the same broad spectrum of beneficial effects that have been observed in rodents and many other short-lived species over the last several decades. Accordingly, excitement about relevance and applicability of DR to retardation of human aging must be tempered with a healthy dose of skepticism when interpreting these results.
Received February 18, 2004
Accepted February 23, 2004
 |
References |
|---|
 Top References |
|---|
| ||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
