Scientific Decision Making, Policy Decisions, and the Obesity Pandemic

Abstract

Rising and epidemic rates of obesity in many parts of the world are leading to increased suffering and economic stress from diverting health care resources to treating a variety of serious, but preventable, chronic diseases etiologically linked to obesity, particularly type 2 diabetes mellitus and cardiovascular diseases. Despite decades of research into the causes of the obesity pandemic, we seem to be no nearer to a solution now than when the rise in body weights was first chronicled decades ago. The case is made that impediments to a clear understanding of the nature of the problem occur at many levels. These obstacles begin with defining obesity and include lax application of scientific standards of review, tenuous assumption making, flawed measurement and other methods, constrained discourse limiting examination of alternative explanations of cause, and policies that determine funding priorities. These issues constrain creativity and stifle expansive thinking that could otherwise advance the field in preventing and treating obesity and its complications. Suggestions are made to create a climate of open exchange of ideas and redirection of policies that can remove the barriers that prevent us from making material progress in solving a pressing major public health problem of the early 21st century.

Obesity represents a complex set of medical, public health, social, and economic problems that have been refractory to study and effective problem solving. Herein we outline a variety of issues that lie at the heart of our inability to understand the fundamental nature of the problem and that impede inquiry aimed at developing effective solutions.

ISSUES AND CONCERNS

Obesity Is a Pandemic

Many developed countries and more affluent sectors of emerging economic powers are experiencing a marked increase in the prevalence of overweight and obesity.^1,2 The implications of this epidemic for human health, productivity, and health care costs are ominous.^3,4

In response to the unfolding crisis, funding of obesity research has increased. For example, the US National Institutes of Health annual funding of obesity research is now nearly $1 billion,⁵ and the combination of nutrition and obesity research has grown to more than $2 billion per year. Despite this outpouring of resources, the listing of 233 categories of research funding does not include a category for physical activity (PA) or exercise.⁶ Although this research investment has led to some advances in understanding the genetic, physiologic, and behavioral correlates of obesity, our ability to treat obesity has increased only modestly outside of surgical interventions and potentially a few newly approved pharmacologic agents, and our ability to prevent obesity at the population level has yet to be demonstrated.⁷ Prevalence in the general population remains very high, and the recent flattening of rates in the United States is not, by generally accepted scientific standards, attributable to any public health initiatives.^8–14

We propose that greater success in this domain may be achieved by overcoming 3 overarching and, to some extent, intersecting shortcomings, each with policy implications: (1) a lack of conceptual clarity in defining obesity, (2) a range of methodologic issues that encourage generating data that perpetuate misconceptions about obesity’s causes, and (3) scholarly dialogues, including peer review processes, that uncritically accept a priori assumptions about cause derived from inaccurate models of obesity and inadequate evidential foundations (Tables 1 and 2). Each of these overarching flaws is associated with a set of policy directives and ad hoc decisions that have contributed to dampening interest in plausible alternatives to conventional explanations as to the causes of and solutions to the obesity epidemic.

TABLE 1.

The Obesity Pandemic: Major Issues in Scientific and Policy Decision Making

Lack of conceptual clarity

The conflation of the definition of obesity and classification of the obese phenotype Definitions of obesity that vary by discipline Epidemiologic and medical: BMI or waist circumference Physiologic: body fat mass or relative body fat percentage

Imprecise measurement

The use of BMI at the population level, although imprecise, is above all practical and acceptable. However, its use at the individual level in obesity interventions is inherently misleading owing to the conflation of body mass with fat mass The storage of energy occurs at the tissue level, yet the measurement and analyses of changes in body mass, fat mass, and adiposity in people examine changes that sum the effect of tissue-level changes across the entire organism and, therefore, cannot distinguish these specific effects

Obesity paradox

Confirmation biases plague research in which the a priori assumption is that obesity is on the causal pathway to many health outcomes (eg, CVD, diabetes) This has led to paradoxes in which, for example, individuals with higher BMIs have better prognoses (eg, cancer, CVD)

Inferring causality

The FLT posits that imbalances between energy intake and energy expenditure invariably lead to alterations in energy storage. Nevertheless, the imprecision of many current measurement protocols renders energy balance a conceptual trap rather than a practical model for understanding obesity and pathologic abnormalities in energy storage Simplistic notions derived from the FLT have led to numerous naive speculations regarding the obesity pandemic and interventions that ignore physiologic and behavioral compensation (eg, physical education or removal of vending machines will decrease obesity)

Responsibilities of the scientific community

The peer review process has failed to curtail the publication of speculations that are conceptually tenuous or devoid of empirical support. For example, if increased food availability was a sufficient cause of obesity, entire populations in developed nations would be obese Lack of due diligence by the research community has facilitated the publication of studies in which vague and imprecise methods have led to results that are suggestive of multiple and divergent explanations

BMI = body mass index; CVD = cardiovascular disease; FLT = first law of thermodynamics.

TABLE 2.

Propagation of Unproven or Even Disproven Propositions as Though They Are Facts^a,^b

Factor facilitating propagation of presumptions and myths	Example of presumptions erroneously accepted as fact that seem to be facilitated by the factor in column 1	Example of myths erroneously accepted as fact that seem to be facilitated by the factor in column 1
Failure to conduct the probative studies (typically RCTs with weight or body fat as outcomes and sufficient durations to test postulated effects on body composition) needed to prove or disprove hypothesized effects	Regularly eating (vs skipping) breakfast is protective against obesity	NA; things typically can become myths only after probative studies are done
A resistance to abandoning cherished ideas even when refuting data exist (ie, cognitive dissonance)	Early childhood is the period during which we learn exercise and eating habits that influence our weight throughout life	Breastfeeding is protective against obesity for the breastfed offspring
Excessive repetitive presentation of the idea (often on the basis of gratuitous association studies long after they are needed) to the point at which the idea comes to be believed merely by the excessive repetition (ie, the mere exposure effect)	Eating more fruits and vegetables will result in weight loss or less weight gain, regardless of whether one intentionally makes any other behavioral or environmental changes	Large, rapid weight loss is associated with poorer long-term weight outcomes than is slow, gradual weight loss
Failure to ask for supporting data from some ideas that seem intuitively obvious	Preventing obesity is easier than treating obesity15	Assessing the stage of change or diet readiness is important in helping patients who seek weight-loss treatment AND
		Setting realistic goals in obesity treatment is important because otherwise patients will become frustrated and lose less weight
Failure to take into account the passive compensation in energy expenditure that occurs merely as a function of changes in body size that occur with initial alterations in energy intake or expenditure	NA; the passive compensation is an established fact, so any calculations that do not accommodate it are effectively myths, not presumptions	Small, sustained changes in energy intake or expenditure will produce large, long-term weight changes that accumulate indefinitely, which fails to take into account compensatory mechanisms
Failure to take into account the active compensation in components of energy balance that may occur when another component of energy balance is manipulated	Snacking contributes to weight gain and obesity	Physical education classes in their current format play an important role in preventing or reducing childhood obesity
Failure to take into account that the active compensation (or lack thereof) in components of energy expenditure that occurs when another component of energy balance is manipulated over a short period (eg, 1 d) may not be maintained if the manipulation is maintained over a longer period; ie, there may be learned compensation	Reducing portion size in some sources of food will reduce body weight or lead to less weight gain in the long-term16	NA
Distortion of the existing scientific evidence by publication and reporting biases or misleading statements	Regularly eating (vs skipping) breakfast is protective against obesity17	Breastfeeding is protective against obesity for the breastfed offspring18

^aNA = not applicable; RCT = randomized controlled trial.

^bData from N Engl J Med¹⁹ References are given only for ideas we classified as myths or presumptions that were not covered by Casazza et al.¹⁹

Major Presumptions and Myths About Obesity

Recently, several articles^19–23 have pointed out that there are many unproven beliefs (presumptions) and disproven beliefs (myths) about obesity erroneously circulating as facts not only in the mass media and the general public but also in the scientific community. Furthermore, these articles suggest some of the factors that may lead to the propagation of these erroneous beliefs. Table 2 provides a nonexhaustive list of examples, quoted or adapted largely from the article by Casazza et al.¹⁹

Defining Obesity: Need for Conceptual Clarity

Obesity is defined differently in various disciplines and, with these variations in definition, individuals, therefore, will be classified differentially. For example, physiologists and researchers working with animal models will classify a subject as obese in terms of absolute or relative (percentage) body fat (BF). Indeed, among individuals in this group, the terms obesity and adiposity are used interchangeably. However, following the lead of public health practitioners, clinicians, some clinical researchers, and epidemiologists, obesity is typically categorized on the basis of a body mass index (BMI [calculated as the weight in kilograms divided by the height in meters squared]) of at least 30.²⁴ Indeed, when obesity statistics are cited, invariably they are based on the convention of defining individuals with a BMI of atleast 30 as obese. On the basis of relatively easily obtained measurements of height and weight, BMI has been used for many decades by clinicians, epidemiologists, and actuaries as a simple estimate of relative weight that can predict health-related outcomes.^25–29 Although it is assumed that the single universal attribute of obesity, defined as a BMI of 30 or greater, is an excess accumulation of adipose tissue, it is clear that BMI is composed of much more than fat mass.³⁰ Over the decades, this has led to debate on the utility of relying on BMI to describe obesity^31–33 and to enhancements to increase its utility.^34–36

Notwithstanding the problems inherent in equating a BMI of 30 or greater with obesity, and then making an implicit assumption that those categorized as obese have excess adipose tissue, this extreme state of overweight is a heterogeneous array of physiologic and behavioral adaptations to environmental conditions.³⁷ Not all people with high levels of BF are obese by this classification; likewise, some people with very high BMIs may have little fat mass. For example, females tend to have much higher levels of BF than their male counterparts at all levels of BMI³⁸; yet, their obesity rates, as defined by a BMI of at least 30, are not generally much higher than those of males.³⁹ In contrast, some individuals (eg, certain categories of elite athletes) may have little BF but BMIs that define them as obese.⁴⁰

Even in most instances in which adiposity is its main driver, obesity is not a single pathologic condition but rather a sign of underlying primary pathologic abnormalities.^41,42 Clearly, obesity arises from the dynamic interplay of the external environment,^43–45 inclusive of the social milieu, built environment, and food energy availability^46–48; behavioral and developmental processes^49,50; and a variety of genes and epigenetic effects that, in turn, control a myriad of metabolic systems and subsystems that regulate body composition, energy intake (EI) and energy expenditure (EE), and nutrient partitioning.^51–53 Although intrinsic (eg, individual genetic and metabolic) factors may play a role in determining who may become obese,^54,55 the preponderance of epidemiologic and ecological evidence indicates that social and environmental factors that determine energy balance (EB) are the primary causes of the secular shift toward higher body weights.^56,57

Definitional issues notwithstanding, characterizing obesity as a distinct condition suggesting a simple, straightforward, and unvarying etiology has biased all phases of obesity-related discourse. Consequently, it has generated a futile search for a single explanation, thus promoting excessively optimistic and generally ineffective interventions and other misguided attempts in the vain hope for “magic (eg, pharmacologic) bullets.” Although BMI provides a reasonable, if crude, estimate of adiposity at the population level, it has little value as a dependent measure in studies of obesity prevention and treatment because changes in body weight may be indicative of alterations in muscle mass, organ mass, other lean body mass, fat mass, body water, glycogen stores, or combinations of these. Therefore, studies reporting decreases in body weight also may exhibit increments in adiposity (ie, percentage fat mass) and decrements in lean body mass.^58,59

Nevertheless, a recent meta-analysis using BMI to determine overweight and obesity status was performed with data from 97 major epidemiologic studies in nearly 2.9 million people with 270,000 deaths during follow-up.⁶⁰ Although this study found that the entire population of obese individuals had a significantly higher mortality rate than did those with a normal BMI (Hazard ratio [HR] = 1.18 [95% CI, 1.12–1.25] for obesity [all grades combined]), this was not the case with class I obesity (BMI, 30.00–34.99), which did not have higher mortality rates and actually had 5% lower mortality than did the normal BMI group (18.50–24.99), which was not quite significant statistically. However, the lowest mortality rate occurred in the overweight BMI group (25.00–29.99), with 6% lower mortality compared with the normal BMI group. Obese individuals 65 years and older had no higher mortality rates compared with normal-weight persons, and this applied even to those with a BMI of 35.00 or greater. These findings from a large metaanalysis of cohort data call into question what is an optimal BMI in current populations.

Obesity Paradox

To add to the existing confusion and controversy, the obesity paradox has been well described during the past decade.⁶¹ Overweight and obesity, generally described by BMI criteria, are associated with an increased prevalence of almost all cardiovascular disease (CVD) risk factors and of almost all CVDs, including hypertension, coronary heart disease (CHD), heart failure, and atrial fibrillation. Despite this association, considerable evidence now indicates that overweight and obese patients with the same CVD diagnoses have better short- and long-term prognoses than do their leaner counterparts. Although it was theorized that this paradoxical relationship may be explained, in part, by the relatively weak relationship between BMI and BF, recent studies have found the obesity paradox with BMI, BF, and even measures of central obesity.^61–64

Adding to the literature reporting the obesity paradox with BMI, we recently found that the highest mortality rates, at least in patients with CHD, are concentrated in those with the lowest levels of lean body mass, suggesting that this may be as much a lean paradox as it is an obesity paradox.^62,64 Nevertheless, in patients with CHD⁶³ and in those with heart failure,⁶⁵ it seems that physical fitness significantly alters the relationship between adiposity and subsequent prognosis such that only those with low levels of fitness demonstrate this obesity paradox. Patients with high fitness, regardless of their level of adiposity by BMI, BF, and central obesity, have a favorable prognosis.^61,63,65 Although further research is needed to determine optimal levels of body composition in patients with CVD, the present data indicate that the greatest attention needs to be directed toward increasing fitness, which is generally obtained with increases in PA and EE, rather than reducing adiposity per se. With respect to CVD, the obesity paradox reflects the complexity of these relationships.

Scientific Standards, Assumptions, Methods, and Discourse Limiting Examination of Alternative Explanations of Cause Presumption

Most scientific theories and models of obesity are based on the first law of thermodynamics (FLT) and proffer that imbalances between EI and EE lead to changes in energy storage (primarily as lipid). The imprecision of current methods for measuring EI^66–70 and EE^71–73 on a large (ie, public health) scale precludes accurate quantification of the EB equation and thus preclude definitive statements regarding the cause of the obesity epidemic.

Description vs Explanation

The FLT provides a true but inadequately simplistic and inherently tautological description of the energy imbalance that leads to overweight and obesity. The “criteria for judging causality,” developed by Hill^74,75 in the early 1950s to provide a framework for assessing cause in complex systems and popularized in the 1964 Surgeon General’s report Smoking and Health,⁷⁶ has been a hallmark of scientific reasoning aimed at assessing evidence in relation to complex causes for more than a half century. The logical flaw in using the FLT to explain secular changes in body weight is reflected in nonadherence to these criteria for judging causality. Specifically, invoking the FLT to explain secular trends in weight gain fails criteria 3, specificity, and 4, temporality. In addition, because of its inability to control for confounding or to estimate effect modification, the FLT provides little useful information for assessing other criteria (ie, the strength of the association, consistency across different sources of evidence, dose-response, plausibility, coherence, and analogy). To illustrate its inability to offer a meaningful explanation of cause, one should consider the analogy of fuel consumption and automobile driving. The average distance driven by motorists has steadily increased since 1980 (in parallel with obesity rates). At the same time, fuel consumption has increased. Although these data are simply and obviously correlated, on their own they do nothing to answer the question: Why are people now driving greater distances? Similarly, to understand the obesity epidemic, we need to search for potentially modifiable root factors that can be measured and modeled to see how well they fit the criteria for judging causality. We should not be satisfied with tautological statements based on the FLT. The flawed logic that naturally flows from naive appeals to the FLT and EB conjectures (eg, increased caloric intake must lead to increased obesity) has biased research funding decisions and the choice of study designs, operational definitions of variables, choice of measurement methods, and analytic procedures, which all condition higher-level decisions regarding how, when, and where to intervene.

Simplistic Determinism and Reductionism

Reliance on FLT-based models leads to neglecting a potentially large array of complex, dynamic, nonlinear, homeostatic systems (eg, energy metabolism and feeding behavior) in favor of static, deterministic equivalency, an extreme form of reductionism in which the complexity of physiologic features and behavior are reduced to basic physics. More importantly, the manifest assumption of FLT-based models is that energy imbalances, no matter how trivial, lead to linear extrapolations to changes in body mass. Measurement and conceptual issues aside, the fallacy of this a priori assumption is obvious; during the past 50 years, it has been reported repeatedly that biological organisms adapt physiologically, behaviorally, or both to perturbations in EB.^58,77–80 The effects of acute or chronic changes in any component of the EB equation will be damped via compensatory alterations in EI (ie, feeding behavior) or EE through changes in body composition or PA or some combination of these.^80,81 In addition, the wealth of responses available to free-living people may well reduce, eliminate, or even reverse the intended effects of any environmental alteration (ie, an intervention).^82–85 This reality makes policy^82,85,86 and funding⁵ decisions based on deterministic equivalencies unproductive in that they incentivize research that fails to account for the array of recalibrations and adjustments that free-living people undergo.

Measurement and Introspection

Very little of the literature on which we base inferences is from the strongest of study designs: randomized controlled trials (RCTs).⁸⁷ Likewise, measurement methods commonly used to assess EI and EE lack the accuracy, reliability, and precision necessary to evaluate the efficacy of interventions or even to draw conclusions from carefully designed observational studies when RCTs are infeasible.^{64,69,71–73,88} This also was (and is) true of the relationship between tobacco exposure and health outcomes, to which the criteria for judging causality were most famously applied.⁷⁶ However, those relationships were easier to sort out for 3 reasons: (1) tobacco exposure is easier to measure than diet or PA (or certainly both), (2) tobacco is very strongly associated with the conditions with which it was first associated (eg, lung cancer⁸⁹), and (3) the effect of tobacco is not as heavily confounded by other factors (although effect modification is important for conditions such as esophageal squamous cell cancer^90,91).

For diet and PA, true criterion measures (eg, doubly labeled water in carefully controlled settings) rarely exist for practical application in large- or even medium-scale epidemiologic or clinical studies, and even doubly labeled water measurements do not assess an important component of PA: intensity. Consequently, marginal intermethod reliability (between-method correlations of ≈ 0.60, indicating that 64% of variability is unexplained) is accepted as adequate validity.^70,92,93 For a variety of practical reasons, observational and even intervention studies rely on inexpensive, inaccurate, and imprecise self-report methods.^94,95 Construct validation is nearly completely ignored,^96,97 and methods that could triangulate to objective measures of exposure are rarely used.^98,99 This satisfaction with inadequate measurement has stunted obesity research and the field of nutritional epidemiology more generally.^100,101

Although practical, including financial, constraints must be acknowledged, the scientific community should demand that methodological research be undertaken to improve the accuracy of self-report, including identifying sources of error with the aim of developing better techniques for statistical adjustment. For example, doubly labeled water (even with its acknowledged weaknesses) could be used to measure EE in carefully controlled intervention studies and could provide a leverage point for improving self-report instruments, as we have done previously.⁷⁰ There are important precedents, most notably from the genetics field, for investing heavily in improved measurement methods and supporting large studies needed to address the complexity of many of the traits under study in human genetics and gene-environment interactions¹⁰² when the potential benefit is perceived to be high in relation to the probable risk from inaction and inattention.

Misuse of Population-Level Data Sources

Numerous data sources are tangential to EI (eg, losses from food inventory and food waste) but are often elevated to causal explanation. Numerous journals have published articles suggesting an invalid determinism (ie, economic forces are not just associated with, but determine, obesity). Although food supply forces (eg, availability and price) may affect purchase and perhaps utilization, the mere presence, purchase, or even increased consumption of food does not necessarily cause long-term changes in adiposity.^{80,81,103–107}

Lack of Due Diligence in Upholding Rigorous Standards of Scientific Evidence When Seemingly Good Ends Are Pursued

Despite that obesity may not simply be the result of overeating,^80,81,104 numerous journals continue to publish articles suggesting that the superficial economics (eg, mere availability of cheap “junk” food) has caused the obesity epidemic.^108–114 Adherence to assumptions with no factual foundation does little to advance scientific understanding. Obesity is widely portrayed as an evil enemy,^115,116 and it is commonly accepted that when we are at war and perceive ourselves to be “the good guys,” valid means of collecting and analyzing data are put aside in favor of expedience in doing what it takes to win the war. The resulting bias permeates the larger social/environmental context, influencing peer review of publication, grants, and, perhaps more importantly, upstream decisions to forgo more expensive, intensive research that addresses the nuances of the complexity of the obesity epidemic.

DISCUSSION AND RECOMMENDATIONS

Our inability to materially and durably decrease the population prevalence of obesity or adiposity in targeted individuals is noteworthy.^8–14 Interventions aimed at increasing EE (eg, through improved physical education classes, incentivizing the use of health clubs/exercise facilities, and adding walking/cycling paths to decrease automobile use and increase human-powered transportation) or decreasing EI (eg, through weight-loss programs, nutrition education, taxing specific food commodities, and food labeling) seem not to result in long-term weight loss, despite appeals to the FLT^117–122.

Crude models of obesity have constrained research for decades, from study design choices, selection and measurement of variables, operational definitions, development and use of measurement tools, and decisions as to how, when, and where to intervene. The flawed logic, vague concepts, inaccurate and imprecise measurements, and unsubstantiated a priori assumptions regarding causal relationships that have followed from vague and poorly specified models during the past century have engendered policy and funding decisions that have incentivized research that perpetuates and compounds these errors via confirmation bias while constraining research paths incongruent with prevailing assumptions.

Conjectures emanating from studies based on the current models of obesity coupled with policy and funding decisions that narrowly constrain future inquiry have contributed to a failure of the peer review process to deepen understanding of obesity at all levels of inquiry. To understand the scope of the problem, we need to remain aware of the fact that obesity arises from the dynamic interplay of the external environment^43–45 with behavioral and developmental processes^49,50 and genetic and epigenetic factors.^51–53 When conceptualized in this manner, it becomes evident that simple, deterministic statements about the etiology of obesity and narrow interpretation about the scope of fundable research are naive and inherently unscientific.¹⁰⁹

Future research should accommodate the apparent complexity of the obesity problem. Conceptualizing a problem implies specification of how to measure and quantify those things that are hypothesized to be important in its cause (or remediation). Because of the narrow interpretation of the causes of the obesity problem, measurement technologies have tended to represent a limited set of possibilities. In planning how to move forward, it is important to consider broadening our interpretation of measurement devices to include study design as an important subset.¹²³ Although RCTs may be a good choice because of the potentially tight control over extraneous factors, it must be realized that often only 1 or 2 factors can be controlled by design. In addition, it is important to appreciate that just because individuals are randomized to a particular study arm in no way guarantees that baseline factors will be entirely equivalent (or exchangeable), that important effect modifiers or confounders will not change differentially over time, that there will be complete adherence, or that there will not be differential dropout.^124,125 Although these factors would plague even simple pharmacologic interventions, it is reasonable to assume that they might be much more severe in situations in which considerable participant commitment is required and blinding is not possible.

Regardless of the specific design chosen, measurements should include potential adaptive behavioral and physiologic responses to specific changes in EI or EE that may mitigate their expected effects. Usually these adaptive behaviors occur in social contexts that integrate physiologic factors with social cues.^126,127 Understanding motivational issues regarding food choices^128–130 and the willingness and ability to engage in a sufficient PA dose (ie, intensity, frequency, and duration) over time^129,131 also represents an important part of successfully designing studies that will materially improve obesity-related outcomes.

Recognizing that such adaptive behaviors cannot be controlled by design,^132,133 existing methods for collecting this kind of information can be used across a variety of experimental and observational study designs for broad applicability in advancing and deepening our understanding of the causes and treatments of obesity.^{9,19,21,22,60} Such statistical methods are well described and commonly used in epidemiologic research, in which effect modification and potential confounding are ubiquitous concerns in the analysis of real-world data, whether collected in the context of an RCT or observationally.^134–136

It also is important to choose appropriately from the continuum of efficacy and effectiveness study designs.¹³⁷ The use of experimental designs may, indeed, be more appropriate for the study of the causes of obesity, and we would tend to recommend such designs in general.^138,139 However, note that the actual differences in EI vs EE necessary to cause measurable changes in energy stores tend to be relatively small (ie, just a few extra kilocalories per day can lead to a couple of kilograms of weight gain per year).^140,141 The fact that even morbidly obese people are in relatively good balance over extended periods of time underlines people’s adaptive ability.^55,80,142 Therefore, compensatory factors that cannot be incorporated as design features of any proposed study must be controlled analytically. This requires anticipating possible mechanisms and either using existing methods or developing new ones to measure these factors and analyze data that are collected using appropriate assessment methods.

To address the public health dimension of the problem, at least some of these studies must be conducted in the context of how people actually lead their lives. Unlike studies of many pharmaceutical agents, behavioral interventions cannot be blinded from the perspective of the participants.^124,143 Also, there may be ethical issues that either inhibit randomization or cause individuals (eg, the morbidly obese) to refuse to be randomized.^60,144

Regarding assessment of the primary exposures, it seems necessary to improve measurement techniques and curtail the use of inaccurate self-reports of diet and PA without identifying methods of adjustment to improve accuracy. We must increase investments in measuring EI and EE more accurately in largescale surveys and epidemiologic studies. Developing more practical objective measurement methods should be a high priority because widely used self-reports are often subject to large sources of measurement bias.¹⁴⁵ Although developing entirely new methods may be desirable, identifying sources of error and then using these data to develop analytic methods to reduce their influence on drawing erroneous inferences might be a desirable interim goal. Rapid progress in developing objective measures of PA that are feasible for use in large-scale studies of free-living individuals is encouraging.¹⁴⁶ Likewise, work that has been conducted on defining sources of measurement error in relation to an “objective” criterionmeasure (eg, doubly labeled water)^70,147 and then using this information to improve estimation of construct validators (eg, body weight, body composition, and serum lipid levels)^148,149 holds promise for advancing this field of obesity research.

CONCLUSION

Current rates of obesity and related conditions continue to place unrelenting strain on health care resources and to reduce productivity. If our response is to be commensurate with the seriousness of the problem, the scientific community must demand higher standards in efforts to understand obesity’s causes and potential solutions.

Abbreviations and Acronyms

BF

body fat

BMI

body mass index

CHD

coronary heart disease

CVD

cardiovascular disease

EB

energy balance

EE

energy expenditure

EI

energy intake

FLT

first law of thermodynamics

PA

physical activity

RCT

randomized controlled trial