Is it Time to Change the Way We Report and Discuss Weight Loss?

This commentary is written to recommend that we consider modifying the way we express weight change in clinical trials of weight management as well as in the physician's office. In most cases, weight loss is expressed as pounds (kg in other countries) or as a percent of baseline weight. We are proposing that weight loss might be expressed as a “percentage of excess body weight loss” (%EBWL), which could complement the use of actual weight loss. %EBWL provides a better estimate of the amount of weight loss that has been achieved relative to a defined goal level; however, that may be defined. This method is typically used to express weight change after surgical treatment for obesity. Using a standard metric will facilitate comparison across behavioral, medical, and surgical trials for weight loss, and may thus be beneficial to patients, clients, health-care providers, investigators, and policy-makers. Weight loss has been expressed in many different ways over the past 150 years ((1)). In his famous pamphlet titled “Letter on Corpulence Addressed to the Public” published in 1863, William Banting expressed his own weight loss in pounds ((2)). Most others have followed his lead, using either pounds or kilograms of actual weight lost. One problem with this approach is that heavier people tend to lose more weight and thus may appear more successful over short periods of time. Expressing weight loss as a percent of initial body weight will partly correct for this, but will not provide a true assessment of the amount of excess weight a patient or client might expect to lose in relation to a floor or maximal likely weight loss. In 1835, Quetelet introduced what is now known as the BMI as a way of normalizing weight for different heights in his studies of populations ((3)). A century and a half later, the BMI was adopted, internationally, as a way of evaluating whether an individual is overweight or obese ((4),(5)). The advantage of the BMI is that it minimizes the effect of height better than other height/weight relationships, except possibly weight/height ((6)). Changes in BMI have also been used as one method of expressing weight loss. Because the height of adults does not change during weight loss, using BMI units can be confusing and provides no advantages over using weight loss alone. Moreover, reporting weight loss as change in BMI units does not reveal how much of the excess weight has been lost. In 1959, a classic article by Stunkard and McLaren-Hume ((7)) reported on a new method for evaluating weight loss in cohorts. They expressed weight loss as a percentage of patients who lost either 20 or 40 pounds. Because heavier people tend to lose more weight, Trulson ((8)) introduced a somewhat more sophisticated approach in which both initial body weight and an appropriate weight loss for the degree of overweight were included. A third criterion was developed by Jolliffe and Alpert ((9)), who proposed the performance index to measure weight loss relative to anticipated or predicted weight loss: This latter concept gets closer to the idea of expressing weight loss in terms of %EBWL, and would be synonymous if the anticipated weight loss was equivalent to the amount of excess weight. Thus its major limitation is the criterion for the anticipated weight loss. The most sophisticated approach was proposed by Feinstein ((10)), and is called the reduction index. WL = weight loss = WI − WE WI = initial weight WS = surplus weight = WI − WT WT = target weight WE = end-of-treatment weight This index considers both the degree of excess initial weight and actual weight loss. Its number ranges from 0 to 200, with higher indices reflecting a greater weight loss. It is difficult to translate this index for individual patients. When a patient begins a weight-loss program, whether at home, in a clinical environment, or as part of a research trial, there is, for all practical purposes, a floor to the amount of weight she/he can lose. Indeed some clinical trials put stopping rules in place in case too much weight is lost. The reason is obvious—we have certain amounts of lean body mass and fat mass that are essential for life. Moreover, once a person has reached what is commonly recognized as the upper limits for a normal or healthy body weight, there is no public health reason to pursue an even lower body weight. Gallagher et al. ((11)) published data showing the amount of body fat for men and women at three different levels of BMI. The data for two age groups are shown in Table 1. Although the data are curvilinear, the line for men parallels that for women. At each BMI listed, women have about 11–12% more of their total body weight as fat as a man at the same BMI. At a BMI of 25 kg/m2 the percentage body fat in women is 32–35% and in men 20–23%. Except for individuals in vigorous physical training, an increase in BMI above 30 kg/m2 is associated with an increase primarily in body fat. Between a BMI of 25 kg/m2 and 30 kg/m2 the BMI does not predict obesity, i.e., increased body fat, as well as at higher BMI values. Body fat asymptotes at about 52% in women and 56% in men ((6)). In a study of anthropometric predictors of body fat, Larson et al. ((6)) found that BMI predicted percent body fat, whereas weight/height was a better predictor of total fat. Some idea about the composition and magnitude of weight changes that might be expected with overfeeding or energy restriction has come from careful assessments of body composition during overfeeding and weight loss. During weight gain in normal individuals, lean tissue accounts for about one-third and fat for about for two-thirds of the weight gain ((12)). Weight loss on energy-deficient diets has shown a similar relationship ((13)). As noted earlier, the surgical literature already typically reports weight loss as %EBWL. This approach takes the degree of excess weight above a recognized benchmark, such as a BMI of 25 kg/m2 or some other agreed-to level, such as the upper limit of standard life insurance tables. We suggest that it makes more sense to use a BMI of 25 kg/m2 for these calculations, because this value corresponds to the upper limit of normal as recommended by the National Heart, Lung, and Blood Institute ((4)) and World Health Organization ((5)). The %EBWL calculated using a BMI of 25 kg/m2 seems a reasonable suggestion for two reasons. First, it focuses on the amount of weight that needs to be lost to reach the upper limit of normal body weight. Second, it has the effect of expressing the weight loss in larger numerical terms, which provides a better estimate for both patient and health-care provider of the true amount of weight that has been lost in any program. Table 2 shows how this approach works. A BMI of 25 kg/m2 is used as the proposed upper limit of normal, although others could be selected based on ethnicity, age, or other considerations. Excesses from this point are calculated. For example, a 95-kg (209-pound) man with a height of 178 cm (70 inches) has a BMI of 30 kg/m2. His weight at a BMI of 25 kg/m2 would be 79 kg. Thus to reach his goal, this man needs to lose 16 kg (95 − 79 kg). If he loses 8 kg or half of this target weight, he will have actually lost 50% of his excess body weight. This becomes a much more realistic description of the success of treatment than 8.4% (8/95), which is his loss from a baseline of 95 kg. After a BMI loss of 5 kg/m2, the percentage of excess body weight is about twice as large as it would be if it were expressed as a percentage of initial weight lost. This is true at any BMI. At a BMI of 50 kg/m2, a 5-BMI-unit decrease would be a 10% weight loss, but nearly a 20% loss of excess body weight. This is a more optimistic outlook for patient and health-care provider and provides a more realistic evaluation of the amount of potential weight loss that can be achieved. Patients and health-care providers are often at odds as to what constitutes meaningful weight loss ((14)). Most physicians believe that a 10% weight loss will produce improvements in metabolic and cardiac risk factors ((4)). However, a 10% weight loss does not match most patients' (clients') desires. Expressed as excess weight loss, this same 10% now becomes ∼20% and illustrates to clients that they have lost 1/5 of the weight that they might have to lose. Second, this approach has the advantage of providing practitioners with a meaningful and clearly defined target weight. We have proposed a BMI of 25 kg/m2 based on accepted standards of weight representation now widely used around the world. ((4),(5)), but other values might be acceptable for different populations. Table 3 provides a way of converting the patient/physician weight-loss goal in percent of total body weight loss into the %EBWL. For these we have used a BMI of 25 kg/m2 as the floor. Using BMI corrects for differences in height between shorter and taller individuals when expressing percent excess weight lost. It will be straightforward to make conversion tables for computers that display the percent excess weight loss for individuals with a given initial BMI, which may be helpful to the patient and provider. Different anticipated weight losses might also be used, as proposed by Jolliffe and Alpert ((9)). There is at least one drawback to using the %EBWL that needs to be acknowledged. With this approach, more obese subjects are less likely to achieve 100% reduction than less obese individuals. This is the case independent of how “normal weight” is defined. In summary, we propose that using %EBWL as a standard metric for reporting the efficacy of behavioral, medical, and surgical weight-loss treatments would unify what are now different ways of reporting results and thus be superior to the current practice. The use of %EBWL may also increase the motivation of individual patients. We believe that this approach has important benefits for patients, health-care providers, obesity researchers, and those who are consumers of the results of that research. Using %EBWL will also increase the comparability among studies of different modalities of treatment in different patient populations with differing starting weights and BMIs. The authors declared no conflict of interest.