Weight Loss and Mortality

Weight cycling, that is, gaining and losing weight over time, has been linked to excess mortality. The studies reported in the supplement to this issue further evaluated the relation among weight loss, weight cycling, and health. The mortality rate was higher among persons who either lost weight or had weight cycling. The reasons for the weight loss or weight cycling were not determined. The association across varying levels of body mass index (BMI) and the findings that the negative effects of weight cycling are greatest in persons with the lowest BMIs suggest that the weight loss was not all voluntary. Attempts were made to adjust for comorbidity and cigarette smoking. The probability is strong that clinical or subclinical disease, other lifestyle changes, or psychosocial factors such as depression accounted for weight loss, weight cycling, and the increased mortality rate.

Disease appears to be epidemic among persons who are at the lower end of the distribution of many biological variables or in whom levels decrease below a critical point. Treatment of hypertension may result in a reduction in blood pressure to below a critical level, the J point, resulting in an increase rather than a decrease in the risk for heart attack [1]. Lowering the blood cholesterol level has been reported to increase the risk for being murdered, committing suicide, and having a hemorrhagic stroke [2]. A low serum albumin level has been identified as a risk factor for cardiovascular disease and probably cancer [3]. Low levels of antioxidant vitamins in the plasma may increase the risks for both heart attack [4] and cancer [5]. Short stature may be a risk factor for heart attack in men and probably in women as well [6, 7]. If a man loses his hair at a relatively early age, the risk for heart attack may increase [8]. We have even recently been told that the quality of sperm is decreasing and may be linked to testicular cancer [9]. Vasectomy may increase the risk for prostate cancer [10].

We are now confronted with studies that suggest that low body weight and weight loss or weight cycling may increase the risk for death, especially from cardiovascular disease [11, 12]. Studies reported in the supplement to this issue of Annals used data from the National Health and Nutrition Examination Survey (NHANES) follow-up, the Framingham Study, and the Multiple Risk Factor Intervention Trial (MRFIT) to evaluate weight loss, weight cycling, and the risk for cardiovascular, noncardiovascular, and total mortality [13-15]. These studies attempted to control for cigarette smoking and other important covariates. Several important observations were made in these studies.

First, the effects of weight loss or weight cycling on cardiovascular and total mortality are not restricted to persons in the highest BMI category. In fact, the effects seem to be inversely related to BMI, which would suggest that the relation between weight loss or weight cycling and mortality may be due to involuntary, rather than voluntary, weight loss. The effects of weight loss also may be attenuated over time. The Framingham Study found a relation between weight loss and a higher prevalence of clinical disease and, possibly, of subclinical disease. None of the studies could explain why the persons lost weight or had weight cycling. Because some of the results were also found among lifetime nonsmokers, variations in weight loss are not associated only with smoking cessation. In fact, other lifestyle changes probably contributed to weight fluctuations.

Are we to take these studies at face value and encourage corpulence, a lifestyle based on the bigger, the better? Should we require new warning labels on track shoes, fruits and vegetables, and low-calorie desserts because overconsumption may increase the likelihood of weight loss, morbidity, and mortality?

Most of the articles in the supplement that relate the low end of a distribution to disease depended on statistical associations that were generated from data sets usually collected for other reasons. The biological plausibility for many, but not necessarily all, of the results has not been determined and, in many cases, variables necessary to evaluate the associations are missing from the analysis. The reasons for weight loss or weight cycling are probably numerous. The most important issue is whether voluntary weight loss increases the risk for disease. To test such a hypothesis, it is necessary to know the other reasons for weight loss, whether clinical or subclinical disease is present, the patient's other habits, his or her alcohol consumption and cigarette smoking status, whether depression is present, and whether hormonal changes are occurring. How then, should we interpret these studies?

First, not one of the studies was a clinical trial designed specifically to test whether weight loss increased or decreased the risk for disease. Such a study would be needed to truly define the risks and benefits of voluntary weight loss. A strong relation exists between overweight or obesity and major risk factors such as blood pressure, cholesterol level, and blood sugar level, and it has been shown that voluntary weight loss reduces these major risk factors. It remains unclear what the mechanism could be whereby voluntary weight loss would improve risk factors for coronary heart disease and yet increase the risk for coronary heart disease.

Second, no study successfully evaluated why participants gained or lost weight or had weight cycling. These studies did not distinguish between voluntary and involuntary weight loss, and no study assessed the specific determinants of weight loss or weight cycling. Changes in lifestyle, especially changes in eating behavior, cigarette smoking, alcohol consumption, and physical exercise are important determinants of weight change. These lifestyle changes may be due to disease or social behavioral factors and may independently affect morbidity and mortality.

Third, the degree of weight change may be a measure of the activity of a disease or of the specific treatments for disease. Inflammation may be important in the pathogenesis of both atherosclerosis and thrombosis [16]. Cytokine changes associated with disease or inflammation may contribute to weight cycling. Psychiatric disorders such as depression may also be related to weight change and to death from coronary heart disease, especially among women [17].

Fourth, weight changes could be related to metabolic changes that increase the risk for disease. Only a few attempts have been made to test this hypothesis, and the results have been negative [11]. Wing [12] recently concluded that there was no consistent evidence that weight cycling affected total body fat, body fat distribution, or resting energy expenditure.

Fifth, there are problems with the measurement of weight change and weight cycling in these studies. Self-reported weights have been used in some studies (for example, NHANES); that is, participants were asked to recall their maximum weight or the number of weight cycles over their lifetime. Weight variability was assessed during a 6- to 7-year period in the MRFIT study and during a 10-year period in the Framingham Study; in neither study were the weight changes occurring during the rest of a person's life considered. Mortality was assessed 5 to 8 years after these measures of weight change, but there was no discussion of what happened to body weight during this 5- to 8-year period.

Given these limitations, what are the implications of these findings to the practice of medicine? As noted above, these studies do not test the efficacy of voluntary weight loss, especially for overweight persons. Only a few studies have analyzed the data on weight cycling or weight change separately for persons who were thin, of normal weight, or heavy. These studies suggest that the negative effect of weight loss is seen primarily in the lowest tertile of body weight. In MRFIT, for example, the association between weight cycling and mortality was in the lowest tertile of BMI, and in NHANES weight losses of 5% to 14% were protective, not detrimental, in the heaviest men (BMI g 29).

These studies on weight loss clearly reinforce the obviousthat weight loss may be a warning sign of disease and increased risk for death. Involuntary weight loss or decreases in such variables as cholesterol level, blood pressure, hemoglobin concentration, and albumin level are not healthy signs, even when previously high levels return to normal. Unintentional weight loss and gain or weight cycling could be a warning sign to a physician of changes in other habits, especially smoking, alcohol use, and drug abuse; preclinical cyclical disease status; or psychiatric disease, especially depression.

These studies have raised interesting new questions, but they lack the in-depth analysis of data to determine a causal pathway between weight cycling, weight loss, low BMI, and disease. It is unlikely that secondary analysis of similar large data sets will provide any further new insights. Future studies need to focus on a much better interpretation of the determinants of weight loss and weight cycling, especially the reasons for the weight loss and cycling, and then on metabolic and health consequences. Such studies clearly need to distinguish between voluntary and involuntary weight loss. It will be possible, for example, to evaluate the reasons for weight loss in some longitudinal studies by directly questioning persons who have lost weight, as well as to carry out a careful clinical assessment of persons who have lost weight, had weight cycling, or have had a stable weight over time.

Many clinical trials of weight loss in the prevention of such disorders as hypertension, diabetes, and cardiovascular disease are either in progress or have been planned. The effects of weight loss and regain, as metabolic variables, need to be evaluated in these studies.

The effects of voluntary weight loss on end points such as cardiovascular disease and cancer can only be determined in randomized trials. Unfortunately, not a single clinical trial assessed the effects of weight loss on major disease outcomes. Given the strong association of obesity and weight gain with blood pressure, glucose level, cholesterol level, and certain cancers (such as breast and prostate), one would suspect that such a trial would show reduced morbidity and mortality rates. It is possible that the observational studies have identified an important health hazardthat weight loss or weight cycling occurs across all weight groups and that even voluntary weight-loss programs could potentially be hazardous. The importance of weight gain and obesity as a public health problem should not be underestimated [18]. It is likely that mortality among heavy cigarette smokers and alcohol consumers who are thin will be higher than among overweight persons who are nonsmokers and have moderate alcohol consumption [19]. However, this certainly does not mean that being obese is good for you. We certainly have not reached a point where we should recommend that all persons weigh themselves in the morning and quickly consume a candy bar if they have lost a half a pound from the previous day or skip a meal or two if they have gained a half a pound in order to remain weight stable and at the lowest risk for heart disease and death.

• Copyright 2004 by the American College of Physicians