Weight loss causes increased mortality: cons

obesity

reviews

© 2003 The International Association for the Study of Obesity.

obesity

reviews

4

, 9–16

9

Blackwell Science, LtdOxford, UKOBRobesity reviews1467-78812003 The International Association for the Study of Obesity. 4Review Article

Weight loss and mortality: cons D. Yang et al.

Address reprint requests to: Dr DB Allison,

Department of Biostatistics, Ryals Public Health

Building, 1665 University Boulevard, University

of Alabama at Birmingham, UAB Station,

Birmingham, AL 35294-0022, USA. E-mail:

[email protected]

Weight loss causes increased mortality: cons

D. Yang

1

, K. R. Fontaine

2

, C. Wang

1,3

and D. B. Allison

1,3

1

Department of Biostatistics, Section on

Statistical Genetics, University of Alabama at

Birmingham, Birmingham, AL, USA;

2

Division

of Rheumatology, Johns Hopkins University

School of Medicine, Baltimore, MD, USA;

3

Clinical Nutrition Research Center, University

of Alabama at Birmingham, Birmingham, AL,

USA

Received 24 September 2002; revised 4

November 2002; accepted 6 November 2002

Summary

Short-term studies indicate that intentional weight loss (IWL) among obese per-sons significantly improves health variables that are often precursors or markersof chronic diseases (e.g. heart diseases, type-2 diabetes). Hence, it is logical toassume that IWL among obese persons would lead to increased longevity. On thewhole, epidemiological studies, including recent ones that use conservative ana-lytic approaches such as distinguishing between apparently IWL and unintentionalweight loss (UWL), adjusting for potential confounders and excluding apparentlyunhealthy subjects, indicate that apparently IWL appears to neither increase nordecrease mortality rate. However, it is important to note that none of the existingstudies were designed specifically to test the hypothesis that IWL reduces mortalityrate, and given methodological problems, these studies do not provide a satisfac-tory way to address the body mass index (BMI)–mortality question. Severalcontrolled clinical trials suggest that IWL may reduce mortality rate. However,even in these studies, it is important to acknowledge that subjects are randomizedto conditions that produce more or less weight loss and not to distinct levels ofweight loss

per se

. Nevertheless, while we await additional data from betterdesigned studies, given our incomplete knowledge, we conclude that it seems morelikely than not that IWL achieved by medically recommended methods does notincrease and probably decreases mortality rate.

Keywords:

Intentional weight loss, mortality, obesity.

Introduction

For over a century, data have been available indicating thatindividuals with an extremely high value of relative bodyweight or obesity have an increased rate of all-cause mor-tality compared with individuals of average body weights(1). Although questions remain, such as the association ofage with ethnicity and whether excess mortality is causedby extreme thinness, there is consensus from the over-whelming majority of studies arguing that obese indi-viduals are at increased risk of all-cause mortality (2). Itappears that this increased mortality rate among obesepersons is largely, though not exclusively, because of causesof death such as heart disease, certain forms of cancer andcomplications of type-2 diabetes (3). Interestingly, obeseindividuals are even at increased risk of mortality afterserious motor vehicle accidents (4).

It is also well documented that when obese individualsundergo procedures that voluntarily produce weight loss,risk factors for many conditions and diseases improve.That is, there are reductions in blood pressure, improve-ments in glucose metabolism, serum lipid profile and qual-ity of life (5–8). Thus, it is quite reasonable to speculatethat intentional weight loss (IWL) among obese personsmay reduce mortality rate. However, definitive data sup-porting the validity of this speculation are wanting. More-over, there are both data and conceptual reasons tocounter-speculate that IWL does not decrease and mayeven increase mortality rate. However, the data supportingthese counter-speculations are substantially flawed. In thispaper, we briefly review evidence related to the issue ofwhether IWL decreases mortality rate. Although it will bereadily acknowledged that the data in support of bothassumptions are incomplete and equivocal, we believe that

obesity

reviews (2003)

4

, 9–16

10

Weight loss and mortality: cons

D. Yang

et al.

obesity

reviews


obesity

reviews

4

, 9–16

the reasoning we present below will demonstrate that themost reasonable conclusion at present is that IWL amongobese persons does not increase and is likely to decreasemortality rate.

What we can observe

We begin by reviewing what we can observe from empiricalstudies, but it must be noted that there is a critical distinc-tion between what we can observe and the reality we maywish to infer.

Animal studies

Although we cannot expect a one-to-one translation of theeffects observed in laboratory animals to human, animalstudies offer a unique opportunity to address the issues athand. Unfortunately, to our knowledge there is no study inrodents that has allowed animals to become and remainobese well into adulthood and then randomized some ani-mals to a weight loss condition and studied the effects onlongevity. We currently have such a study underway, butresults are not yet available. To date, calorie restriction(CR) is the only intervention identified that has consistentlyprolonged life span and decreased mortality rate in multiplespecies including mammals. Although CR should not beconflated with low body weight, it must be acknowledgedthat CR markedly decreases body weight and fat mass (5).More relevant to human studies is middle-age onset CR.Although the effects of late onset CR on mortality are lessconsistent among studies (6,7), given the possibility thatmiddle-age onset CR not only prevents obesity but alsocauses weight loss, it is plausible to investigate further therole of weight loss in the life-prolonging mechanism ofmiddle-age onset CR.

Association of body mass index with mortality rate

The currently available data clearly demonstrate that theassociation between body mass index (BMI) and mortalityis J- or U-shaped (8). That is, increased mortality rate isobserved at very low and high BMIs. Among white malesand females under the age of about 65 years, the nadirsof the curve tend to be around the mid to high 20s forBMI. Although isolated studies may occasionally showother results, these conclusions are clearly supported bythe overall body of data. This pattern of findings isrobust and generally unchanged when various statisticalapproaches are applied to the data to control the influenceof potential confounders (i.e. excluding smokers, earlydeaths, those with recent weight loss or weight fluctuation)(8,9).

Association of weight loss with health benefits in clinical trials

The effects of obesity on health have been well summarized(10). Although short-term studies have consistently dem-onstrated the benefits of weight loss, there is a dearth ofwell-designed studies that examine the benefits of long-term weight loss. Nevertheless, the few long-term studiesthat have been performed tend to confirm the results ofbriefer studies (11). Specifically, weight loss has been shownto produce considerable benefit with regard to type-2diabetes mellitus and insulin resistance, cardiovascular dis-ease and hypertension, dyslipidaemia, ovarian function,knee osteoarthritis and obstructive sleep apnoea (12,13).Although it is sometimes difficult to disentangle the effectsof weight loss from the effects of lifestyle modification (e.g.healthier diet, increased physical activity), weight loss sub-stantially ameliorates a variety of risk factors and chronicdiseases that are associated with obesity.

Results from well-known obesity trials

In a 9-year study of 233 elderly people (mean age 65 years;range 55–85 years) with newly diagnosed type-2 diabetesrandomized to either dietary counselling or controls, Leanand colleagues (14) found that each kilogram of weight losswas associated with 3 months of increased survival.

In a recent analysis of data from the large-scale SwedishObese Subjects (SOS) study (15), a study involving nearly2000 matched pairs of obese adults (men: BMI

=

34;women: BMI

=

38) randomized to either obesity surgery orconventional obesity treatment, it was found that, at2 years, an average weight loss of 28 kg among the surgerygroup was associated with a reduced incidence of diabetesby 32 times compared with the controls. After 8 years, therewas a fivefold reduction in diabetes incidence. With regardto hypertension, the 2-year incidence was reduced by 2.6times in the surgery group. After 8 years, the incidence ofhypertension was nearly equal between the two groups.

Recent results from the Finnish Diabetes PreventionStudy suggest that lifestyle modification (i.e. change in dietand increased physical activity) that produces modestweight loss reduces the risk of type 2 diabetes (16). Specif-ically, 522 middle-aged overweight men and women withimpaired glucose tolerance were randomized either to a dietand exercise protocol or to a control group. The interven-tion produced a 4.7% loss of baseline weight after 1 year.After 3 years, the intervention group had a 58% reducedrisk of diabetes compared with the controls. The 4-yearcumulative incidence of diabetes was 11% for the interven-tion group and 23% for the controls.

Recently, the Diabetes Prevention Program reportedresults from a similar intervention carried out in the USA(17). Mean net weight losses of 7%, 6% and 4% over 1,

obesity

reviews


D. Yang

et al.

11


obesity

reviews

4

, 9–16

2 and 4 years, respectively, were associated with a 58%lower diabetes incidence rate than participants in thecontrol group. In a year-long randomized controlled trial(RCT) of 406 coronary heart disease patients randomizedto dietary intervention or controls, Singh

et al

. (18) foundthat a 6% weight loss among those in the interventiongroup was associated with a 40% reduction in cardiacevents and a 45% reduction in mortality compared withthe controls.

Collectively, these long-term RCTs suggest strongly thatIWL significantly improves health parameters that are asso-ciated with several obesity-related comorbidities.

Outcomes of surgically induced weight loss

The method that is most effective at producing substantiallong-term weight loss is surgery (19). To date, no large,randomized, well-controlled clinical trials of obesity sur-gery with respect to efficacy in terms of mortality rate havebeen published. However, Drenick

et al

. (20) conducted anobservational study among a group of 200 morbidly obesemen who were followed for a mean period of 7.5 years.About 25% of the men in the study died during the courseof the study. According to Drenick

et al

., ‘life-table tech-niques comparing the mortality among the obese with thatamong men in the general population, demonstrated a12-fold excess mortality in the obese in the age group 25–34 years, and a six-fold excess in the age group of 35–44 years.’ In contrast, Benotti

et al

. (21) reported on 5178patients with morbid obesity who underwent surgical treat-ment in 12 different centres, and stated that ‘operative andlate mortality rates were quite similar to observed deathrates for non-obese men and women between 25–64 yearsof age. These data suggest that gastric surgery for morbidobesity results in significant reduction of health risk.’ Thus,though hardly enjoying the inferential strength of well-conducted, randomized clinical trials, the combination ofobservational data among morbidly obese individuals andmortality rates observed among morbidly obese individualswho have undergone surgical treatment suggests that sur-gical treatment may have a profound beneficial effect onoverall mortality rate.

The ongoing SOS study, a longitudinal randomized trialinvolving obesity surgery vs. conventional obesity treat-ment, should offer a rigorous evaluation of the effects ofsurgically induced weight reduction on both the incidenceof obesity-related comorbidities and mortality.

Association of weight loss and mortality rate in epidemiological studies

The results of the observational studies, including thosewhere the investigators adjust for potential covariates,exclude unhealthy subjects or distinguish between IWL and

unintentional weight loss (UWL), indicate that weight lossdoes not appear to either increase or decrease longevity(22). Thus, regardless of how the data are ‘cleaned’ ortreated statistically, weight loss does not appear to eitherincrease or decrease longevity. This literature, however, isplagued with methodological concerns including (i) thedifficulty of assessing and accurately taking account ofweight regain after weight loss; (ii) the need to determinethe method(s) by which the weight loss was achieved; (iii)assessing dietary changes during follow-up; and (iv) the useof BMI (kg m

-

2

) as a proxy for adiposity.

Effect of body composition

Epidemiological studies have, by and large, used BMI as aproxy for adiposity. However, BMI may not provide thebest estimate of the impact of weight loss on mortalitybecause BMI is composed of two components, fat mass(FM) and fat-free mass (FFM). Thus, the use of BMI maymask any differential health consequence associated withthe loss of both FM and FFM (Table 1). In this regard,Allison

et al

. (23) investigated the differential effects of FMand FFM on all-cause mortality rate in a nationally repre-sentative sample of over 10 000 men. Using anthropomet-ric data, skin folds, from the first and second NationalHealth and Nutrition Examination Surveys follow-up stud-ies, they observed that, as predicted, BMI had a U-shapedassociation with mortality. However, when measures of FMand FFM were added to the model, the BMI–mortalityassociation became nearly monotonically increasing andthe FM–mortality association was monotonically increas-ing, while the FFM–mortality association was monotoni-cally decreasing. These results show that the use of BMI asa proxy for adiposity may mask differential associationsbetween body composition and mortality, suggestingstrongly that future studies should obtain not only BMI butalso direct measures (e.g. dual-energy X-ray absorptiome-try scans) of body composition. Moreover, if confirmed,results such as these suggest that weight loss should berecommended only under conditions where a sufficient pro-portion of the weight lost would be FM. Allison

et al

. (24)also studied the independent effects of weight loss and fatloss on mortality rate and found that, among individualswho are not severely obese, weight loss is associated withincreased mortality rate while fat loss with decreasedmortality rate.

Unfortunately, with the exception of increased aerobicand resistance exercise in conjunction with a low-caloriediet, the means to produce such a weight loss remainunknown (25,26). However, research on animals is under-way to investigate pharmacological compounds that mayinfluence the proportion of weight lost as fat as opposedto lean tissue. This should assist in the development ofpharmacological agents that produce ‘targeted’ weight loss

12


D. Yang

et al.

obesity

reviews


obesity

reviews

4

, 9–16

in humans. It also remains to be determined whether losingabdominal fat, as opposed to fat deposits in other regions,associates with longevity. Given the well-documentedincreased health risks associated with abdominal fat (27),it is likely that reducing abdominal fat has beneficial effectswith regard to longevity.

Effect of the ‘so-called intentionality’

The issue of intentionality of weight loss is critical. Forexample, we recently reported (9) that, using the attribut-able risk approach on published data on the attribution ofweight-loss intention in a large population-based survey(28), among the overweight women who intended to loseweight, 46% did so but 19% would have lost weight evenif they did not intend to do so. Thus, the fraction of weightloss among these women as a result of factors other thantheir intention is about 41% (19/46). This suggests that asizeable subportion of those who have been classified asintentional weight losers in past studies may have lostweight for other reasons (e.g. subclinical disease) and high-lights the limitations inherent in trying to determine theeffect of IWL on mortality using observational data.

Nevertheless, most of the available information on theassociation between so-called IWL and mortality is derivedfrom observational studies [see ref. 22 for a recent review].However, because of the inherent limitations in thesestudies, it seems clear that other sources of information areneeded. To date, these include non-RCTs of weight-lossinterventions and RCTs of lifestyle interventions that hadweight loss as a secondary outcome. Unfortunately, fewRCTs have focused specifically on the effects of weight losson morbidity or mortality. Those that have been conductedhave tended to focus on two known comorbidities of obe-sity: hypertension and diabetes (29).

In a recent comprehensive review of these studies, Gregg& Williamson (29) conclude that (i) IWL produces many

short-term benefits, especially with regard to reduced riskfactors of cardiovascular disease and diabetes; (ii) IWLdoes not increase and may decrease mortality amongpeople with obesity-related health conditions; and (iii) life-style interventions that promote modest IWL can preventdiabetes and hypertension among those at high risk of thesediseases.

Shadows on the wall: conceptual issues

As in Plato’s

Allegory of The Cave

, the challenge we facein studying the relation between weight loss and mortalityrate is that we only have the opportunity to observe theshadows on the wall and, from these distantly relatedobservations, we must draw conclusions about the truephenomena casting the shadows. That is, we are interestedin what the data mean rather than the data

per se

. Thisrequires considering the nature of the phenomena thatcould be giving rise to the observed patterns of data. In thisregard, a number of conceptual issues must be considered.

Confounding

It is well recognized that confounding may occur whenexamining the relationship between weight loss and mor-tality rate. For example, people who quit smoking typicallygain weight. Smoking cessation is likely to lead to increasedlongevity. Thus, if studies did not account for changes insmoking rates, this might confound the observed associa-tions. However, our review of the literature suggests thatsmoking status and any change in smoking status is typi-cally accounted for in weight loss–mortality studies (22).The most frequently cited confounding variable of interestis occult or subclinical disease. That is, some individualsmaybe suffering from an underlying disease process thatis both causing weight loss and leading to an increasedmortality rate. This is difficult, if not impossible, to fully

Table 1

Selection of recent studies evaluating the effects of fat mass (FM) and fat-free mass (FFM) on mortality rate (MR)

Authors Sample Measures of FM & FFM Effects on MR

Charles

et al

. (35) 7608 men from the Paris ProspectiveStudy

Anthropometry (skin folds) FFM appeared protective

Baik

et al

. (36) 39 756 men from the Health Professional Follow-up Study

Anthropometry (skin folds) Waist circumference positively associatedwith CVD mortality among men

≥

65 yearsHeitmann

et al

. (37) 787 men from Goteborg, Sweden born in 1913

Anthropometry (skin folds) FM monotonically increasing MR; FFM monotonically decreasing MR

Lahmann

et al

. (38) 28 098 adults from the Malmo Dietand Cancer Prospective CohortStudy

Bio-impedance %FM increased MR in men but not women;FFM unrelated to MR

Allison

et al

. (23) 25 318 adults from NHANES I andNHANES II follow-up studies

Anthropometry (skin folds) Among men: FM monotonically increasing MR;FFM monotonically decreasing MR

Adapted from Fontaine & Allison (8). NHANES, National Health and Nutrition Examination surveys; CVD, cardiovascular disease.

obesity

reviews


D. Yang

et al.

13


obesity

reviews

4

, 9–16

control for because occult disease is, by definition, unob-servable. Investigators have tried to control for this possi-bility by categorizing people into those experiencing IWLand those experiencing UWL. However, as we will discussin the subsequent sections on measurement, the constructvalidity of an indicator of someone’s intention to loseweight as a measure of the degree to which their weightloss is because of subclinical or occult disease is highlyquestionable.

Still other types of confounding may occur. One possi-bility is the so-called ‘confounding by indication’

.

Con-founding by indication refers to the possibility thatindividuals who have some indication that they are at riskof experiencing the events in question (i.e. early mortalityin this case) change their behaviour with respect to the riskfactor under study. Thus, for example, people who have aparticular genotype that makes them more susceptible tothe ill effects of obesity-related diseases, such as diabetes,may observe a parent or sibling die early from such condi-tion(s). These people, recognizing their own susceptibility,may then make greater efforts to lose weight than an indi-vidual who has not had this experience. This would leadto greater weight loss among obese people being correlatedwith greater genetic susceptibility to the ill effects of obesity.This could again make the effects of weight loss appear lessbeneficial or more deleterious than they are in reality. Obvi-ously, many other types of confounding can occur. The listof possible confounders is limited only by one’s imagination.

The only way to definitively eliminate confounders is torandomize subjects to levels of the independent variableunder study. This is true because randomization is the onlymethod that controls for both known and unknown andmeasurable and unmeasurable confounders. Unfortunately,it does not seem possible to randomize people to differentdegrees of weight loss. Rather, one can, at best, randomizepeople to different treatment conditions that are associatedwith different degrees of weight loss. Although this maybring us closer to the experimental design we would ideallywish for, it must be recognized that within either experi-mental condition subjects are assigned to, the weight lossthey achieve will still be one that is a potential mixture ofweight loss which is because of the treatment itself andfactors other than treatment. That is, subjects are not ran-domized to varying degrees of weight loss. Rather, they arerandomized to treatments that produce varying degrees ofweight loss. Thus, while we can rely on the inferentialpower of randomization to determine the effects of weight-loss treatments on mortality rate, we cannot rely on theinferential power of randomization to determine the effectsof weight loss

per se

on mortality rate.Therefore, the strict inference one can achieve in ran-

domized clinical trials of treatment effects is not possiblewhen examining effects of weight loss on mortality rate.This means that we will always be left in the position of

reviewing the possible role of confounding variables byargument rather than definitive demonstration. It is ourcontention that the potential for confounding, even in thebest studies conducted to date, is still high and, therefore,we are unconvinced by the presently available data thatIWL increases mortality rate.

Measurement

In interpreting the relationship between a risk factor andan outcome, it is important to consider the measurementquality of both variables. Again, what we observe is notthe association between the risk factor and the outcomebut rather the association between our measurement of therisk factors and our measurement of the outcome. If eithermeasurement is lacking in reliability or validity, the resultsobtained are open to question. In the case of mortality, itis generally presumed that we measure the time of deathwith sufficient reliability and validity to be unconcernedabout this aspect, and we do not dispute this point. Withrespect to weight loss, regardless of intention, we alsobelieve that it can be measured with great reliability. Spe-cifically, if the weight loss in a study is based upon twomeasures of weight obtained on a calibrated scale (ratherthan self-report), we would have little concerns about thereliability. However, we do have marked concerns aboutthe validity of the observed weight change as an indicatorof IWL. To date, IWL has been defined as the degree ofweight change that is observed in people who state at thebeginning of some interval that they are trying to loseweight. We believe that this measure of IWL is of highlyquestionable validity. This is because the weight changethat an individual experiences after having stated an inten-tion to lose weight is likely to be a mixture of weightchange that is because of their intention (or more accu-rately, because of the behaviours they exhibit in responseto their intention) and weight change that is because offactors beyond their intention. That is, just as individualswho intend to lose weight and subsequently do so may haveexperienced that weight change because of multiple factors,including stress and disease, individuals who state that theyintend to lose weight may also lose weight because ofpotential stresses and diseases. Therefore, models are nec-essary that allow us to disentangle the component of weightloss because of intention and that because of factorsbeyond one’s intention. We are currently developing suchmodels. Nevertheless, even these models must be basedupon assumptions about unobservable quantities. There-fore, even after they are available, questions will always beraised about our ability to actually observe the effects ofIWL. In this context, the more important point is that theobserved weight change that has occurred among peoplewho stated an intention to lose weight which has, hereto-fore, been used as a proxy for IWL is, in our opinion, an

14


D. Yang

et al.

obesity

reviews


obesity

reviews

4

, 9–16

unacceptable proxy for IWL and may lead to markedlymisleading results. This was illustrated by analysis of thedata from Meltzer & Everhart (28).

Body composition vs. body weight

As indicated earlier, there is a burgeoning body of evidencewhich suggests that changes in body weight are less impor-tant determinants of health and longevity than are changesin body composition (Table 1). Body weight and changestherein reflect a mixture of different body mass componentsand changes therein, respectively. Failure to examine thesecomponents separately may lead to spurious conclusions.Just as we no longer examine total cholesterol as a riskfactor of heart disease having recognized that it is com-posed of multiple components, at least two of which(low- and high-density lipoprotein) have markedly differenteffects in opposite directions; lumping these two compo-nents together makes little sense. The same can be said ofFM and FFM. Indeed, it is noteworthy to recall that obesityis generally conceptually defined as an excess of body fat.We only operationally define it by an excess of body weight(i.e. a high BMI). From this perspective, the question ofwhether IWL decreases or increases mortality rate maybethe wrong question. The right question maybe ‘Does reduc-tion in adiposity decrease mortality rate?’ If and when weare convinced that we have obtained an affirmative answerto that latter question, we will then be poised to ask otherpertinent questions such as ‘Do we have the ability toselectively promote fat loss and, if so, how?’

Role of diet quality

In the animal CR literature, it has been observed thatgreater longevity is achieved when animals are subjected towhat has been called ‘under-nutrition without malnutri-tion’ (5). Specifically, when the total amount of caloriescalorically restricted animals are allowed to consume isreduced, the vitamin and mineral mix that they are givenis generally increased to compensate for the reduction intotal calories. This ensures that their micronutrient intakeremains adequate. In contrast, it is well-documented thatmany people do not have adequate micronutrient intakes(30). Assuming that diet composition does not radicallychange when individuals lose weight in epidemiologicalstudies, then such individuals are likely to have furtherreduced their micronutrient intake below acceptable levelson average. This may account for some of the apparentlydeleterious effects of weight loss on longevity as well as theapparently deleterious effects of extreme thinness on lon-gevity. In contrast, it has been shown that, among individ-uals undergoing voluntary weight loss in clinical trials,dietary quality typically improves (31). To the extent thatdiet quality and micronutrient intake in particular are

important variables, only by carefully examining controlledclinical trials where micronutrient intake is monitored willwe be able to adequately study the effects of IWL withoutconfounding from poor micronutrient intake.

What are the real models of interest?

Once again, we wish to point out that the question of whatis the effect of IWL may not really be the important ques-tion of interest. A more important question maybe some-thing along the lines of ‘What is the effect of weight lossachieved by a method recommended by modern medicalpractice as safe and effective?’ We have ample data indicat-ing that when people try to lose weight, they do so througha wide variety of methods, including CR, increased activity(which can take on many different types and degrees ofvigour), legal and illegal drugs, cigarettes, dietary supple-ments, self-induced vomiting and multiple other methodsof dubious efficacy and safety (31). To take perhaps themost extreme example, it is well-documented that youngwomen often use cigarettes as a means of weight control(32). It should be hardly surprising that, among such peo-ple, successful attempts to lose weight would be associatedwith increased mortality rate when the weight loss isachieved by increasing cigarette smoking. In contrast, if onedoes not intend to lose weight but is exposed to newenvironmental circumstances that lead to decreased con-sumption of saturated fat and calorically dense foods,increased consumption of fibre, fresh fruits, fresh vegeta-bles and increased performance of moderate physical activ-ity on a regular basis, this would probably lead to weightloss. We would expect such weight loss to be beneficial,even if not intentional. This highlights that categorizingpeople simply into intentional and unintentional categoriesmay miss the point. Future studies would be better offcategorizing people on the basis of the methods by whichtheir weight loss is achieved.

Even if an individual intentionally seeks weight loss, it isimportant to note that many weight-loss practices currentlyin use (e.g. herbal supplements, fad diets) are of unknownefficacy and safety (33). Hence, it appears ill advised toestimate the effects of weight loss achieved by medicallyrecommended methods by studying weight loss that ismerely reported to be intentional. What we need are studiesof weight loss among obese humans that is achieved bymodern methods accepted by mainstream medicine.

New data needed and on the horizon

We have argued and presented data which suggest that:

• The currently available data are quite imperfect andincomplete with respect to their ability to inform us aboutthe true effects of IWL.

obesity

reviews


D. Yang

et al.

15


obesity

reviews

4

, 9–16

• The true effects of IWL on mortality rate are plausiblynot deleterious.

• The balance of evidence can be interpreted as suggest-ing that weight loss achieved intentionally by medicallyrecommended methods among obese people is likely toresult in decreased mortality rate. However, in order toconsolidate our knowledge on this topic, much workremains to be done.

Epidemiological data

New data from the field of epidemiology are forthcoming.Our own group is attempting to initiate studies examiningthe effects of rigorously measured body composition onmortality rate in humans. A large, ongoing epidemiologicalstudy referred to as ‘Health ABC’ will collect similar data(34). Finally, we need epidemiological investigators to con-duct much more rigorous assessments of current healthstatus and how and why people change weight so that wecan evaluate the differential effects of various methods ofweight reduction on mortality rate.

Controlled clinical trials

There are also at least two large controlled clinical trialsunderway that may prove informative in this regard. Thefirst is the SOS study described in the section ‘Results formwell-known obesity trials’. The SOS study involves severalthousand morbidly obese subjects who are either undergo-ing obesity treatment through surgery which produces, onaverage, approximately a 16% reduction in body weight at5 years after treatment, and a control group of subjectswho receive usual care and lose little, if any, weight. Mor-tality data are not available but should be forthcoming.Second, an ongoing study, sponsored by the NationalInstitutes of Health, called ‘Look AHEAD’ involvesapproximately 6000 obese individuals with diabetes. Thismulticentre, randomized clinical trial will compare avigorous weight-loss programme to an appropriate controlgroup. Although, as stated earlier, this is not equivalent torandomizing subjects to weight loss

per se

, it should becapable of providing us with information about what theeffects of weight-loss treatments (but not necessarily weightloss itself) are on mortality rate.

Conclusion

On balance, epidemiological studies indicate that IWLappears to neither increase nor decrease mortality rate.However, none of the existing studies were designed spe-cifically to test the hypothesis that IWL reduces mortalityrate, and given the methodological problems we have out-lined above, these studies do not provide the appropriatedata to derive a satisfactory answer to the BMI–mortality

question. As the design and measurement techniques usedin observational studies become more sophisticated, includ-ing serial measures of body composition, we maybe in abetter position to draw valid conclusions of the effect ofIWL on mortality.

In contrast to the epidemiological data, several short-and longer-term controlled clinical trials suggest that IWLmay reduce mortality rate. However, it is unlikely thatthese clinical trials will offer a definitive answer to theBMI–mortality question, because even in these studies sub-jects are randomized to conditions that produce more orless weight loss and not to distinct levels of weight loss

per se

.Thus, in the interim, while we await additional data from

better designed studies, it seems premature to conclude thatIWL either increases or decreases mortality rate. However,based on the preponderance of the current evidence, whileweighing its inherent weaknesses, we conclude that it ismore likely than not that IWL achieved by medicallyrecommended methods does not increase and probablydecreases mortality rate.

Acknowledgements

This work was supported in part by NIH grantsR01DK054298 & P30DK56336 and a grant from theArthritis Foundation.

References

1. Kahn

H, Williamson

DF. Abdominal obesity in 19th centuryAmerican men.

Int J Obes Rel Metab Disor

1994;

18

: 686–691.2. Allison

DB, Fontaine

KR, Manson

J, Stevens

J, Van Itallie

TB.How many deaths are attributable to obesity in the United States.

JAMA

1999;

282

: 1530–1538.3. Allison

DB, Saunders

SE. Obesity in North America. An over-view.

Med Clin North Am

2000;

84

: 305–332.4. Mock

CN, Grossman

DC, Kaufman

RP, Mack

CD, Rivara

FP.The relationship between body weight and risk of death or seriousinjury in motor vehicle crashes.

Accid Anal Prev

2002;

34

: 221–228.5. Weindruch

R, Sohal

RS. Seminars in medicine of the Beth IsraelDeaconess Medical Center. Caloric intake and aging.

N Engl JMed

1997;

337

: 986–994.6. Murtagh-Mark

CM, Reiser KM, Harris R Jr, McDonald RB.Source of dietary carbohydrate affects life span of Fischer 344 ratsindependent of caloric restriction. J Gerontol A Biol Sci Med Sci1995; 50: B148–B154.7. Goodrick CL, Ingram DK, Reynolds MA, Freeman JR, CiderN. Effects of intermittent feeding upon body weight and lifespanin inbred mice: interaction of genotype and age. Mech Ageing Dev1990; 55: 69–87.8. Fontaine KR, Allison DB. Obesity and mortality rates. In:Bouchard C, James WPT, Bray GA (eds). Handbook of Obesity.Marcel Dekker, Inc.: New York, in press.9. Allison DB, Heo M, Fontaine KR, Hoffman D. Body weight,body composition, and longevity. In: Bjorntrop P (ed.). Interna-tional Textbook of Obesity. John Wiley & Sons: Sussex, 2001, pp31–48.

16 Weight loss and mortality: cons D. Yang et al. obesity reviews

© 2003 The International Association for the Study of Obesity. obesity reviews 4, 9–16

10. National Task Force on the Prevention and Treatment ofObesity. Overweight, obesity, and health risk. Arch Intern Med2000; 160: 898–904.11. Pi-Sunyer FX. A review of long-term studies evaluating theefficacy of weight loss in ameliorating disorders associated withobesity. Clin Therapeutics 1996; 18: 1006–1035.12. World Health Organization. Obesity: Preventing and Manag-ing the Global Epidemic. Report of a WHO consultation onobesity. WHO: Geneva, 1998.13. Goldstein DJ. Beneficial health effects of modest weight loss.Int J Obes 1992; 16: 397–415.14. Lean MEJ, Powrie JK, Anderson AS, Garthwaite PH. Obesity,weight loss and prognosis of type 2 diabetes. Diabetic Med 1994;7: 228–233.15. Torgerson JS, Sjostrom L. The Swedish Obese Subjects (SOS)study – rationale and results. Int J Obes Relat Metab Disord 2001;25: S2–S4.16. Tuomilehto J, Lindstrom J, Eriksson JG, Valle TT,Hamalainen H, Ilanne-Parikka P, Keinanen-Kiukaanniemi S,Laakso M, Louheranta A, Rastas M, Salminen V. Prevention oftype 2 diabetes mellitus by changes in lifestyle among subjects withimpaired glucose tolerance. New Engl J Med 2001; 344: 1343–1350.17. Diabetes Prevention Program Research Group. Reduction inthe incidence of type 2 diabetes with lifestyle intervention ormetformin. N Engl J Med 2002; 346: 393–403.18. Singh RB, Rastogi SS, Verma R, Laxmi B, Singh R, Ghosh S,Niaz MA. Randomised controlled trial of cardioprotective diet inpatients with recent acute myocardial infarction: results of one-year follow-up. BMJ 1992; 304: 1015–1019.19. Kral J. Surgical treatment of obesity. In: Bjorntrop P (ed.).International Textbook of Obesity. John Wiley & Sons: Sussex,2001, pp 511–518.20. Drenick EJ, Bale GS, Seltzer F, Johnson DG. Excessive mor-tality and causes of death in morbidly obese men. JAMA 1980;243: 443–445.21. Benotti PN, Hollingshead J, Mascioli EA, Bothe A Jr, BistrianBR, Blackburn GL. Gastric restrictive operations for morbid obe-sity. Am J Surg 1989; 157: 150–155.22. Fontaine KR, Allison DB. Does intentional weight loss affectmortality rate? Eating Behav 2001; 2: 87–95.23. Allison DB, Zhu SK, Plankey M, Faith MS, Heo M. Differ-ential associations of body mass index and adiposity with all-causemortality among men in the first and second National Health andNutrition Examination Surveys (NHANES I and NHANES II)follow-up studies. Int J Obes Relat Metab Disord 2002; 26: 410–416.24. Allison DB, Zannolli R, Faith MS, Heo M, Pietrobelli A,VanItallie TB, Pi-Sunyer FX, Heymsfield SB. Weight loss increasesand fat loss decreases all-cause mortality rate: results from two

independent cohort studies. Int J Obes Relat Metab Disord 1999;23: 603–611.25. Pavlou KN, Steffee WP, Lerman RH, Burrows BA. Effects ofdieting and exercise on lean body mass, oxygen uptake, andstrength. Med Sci Sports Exerc 1985; 17: 466–471.26. Whatley JE, Gillespie WJ, Honig J, Walsh MJ, Blackburn AL,Blackburn GL. Does the amount of endurance exercise in combi-nation with weight training and a very-low-energy diet affectresting metabolic rate and body composition? Am J Clin Nutr1994; 59: 1088–1092.27. Rosmond, R. Visceral obesity and the metabolic syndrome.In: Bjorntorp P (ed.). International Textbook of Obesity. JohnWiley & Sons: New York, 2001, pp 339–350.28. Meltzer AA, Everhart JE. Correlations with self-reportedweight loss in overweight US adults. Obes Res 1996; 4: 479–486.29. Gregg EW, Williamson DF. The relationship of intentionalweight loss to disease incidence and mortality. In: Wadden TA,Stunkard AJ (eds). Handbook of Obesity Treatment. Guilford:New York. 2001, pp 125–143.30. Chakravarty I, Sinha RK. Prevalence of micronutrient defi-ciency-based on results obtained from the national pilot programon control of micronutrient malnutrition. Nutr Rev 2002; 60:S53–S58.31. Neumark-Sztainer D, Sherwood NE, French SA, Jeffery RW.Weight control behaviors among adult men and women: cause forconcern? Obes Res 1999; 7: 179–188.32. Tomeo CA, Field AE, Berkey CS, Colditz GA, Frazier AL.Weight concerns, weight control behaviors, and smoking cessa-tion. Pediatrics 1999; 104: 918–924.33. Allison DB, Fontaine KR, Heshka S, Mentore JL, HeymsfieldSB. Alternative treatments for weight loss: a critical review. CritRevs Food Sci Nutri 2001; 41: 1–40.34. Franse LV, Di Bair M, Shorr RI, Resnick HE, van Ejik JT,Bauer DC, Newman AB, Pahor M. Type 2 diabetes in old well-functioning people: who is diagnosed?: Data from the Health,Aging and Body Composition Study. Diabetes Care 2001; 24:2065–2070.35. Charles MA, Oppert J-M, Thibult N, Guy-Grand B, Ducime-tiere P. Muscle mass, fat mass, and mortality: 15-year follow-upof the Paris Prospective Study. Obes Res 2000; 7: S59.36. Baik I, Ascherio A, Rimm EB, Giovannucci E, Spiegelman D,Stampfer MJ, Willett WC. Adiposity and mortality in men. Am JEpidemiol 2000; 152: 264–271.37. Heitmann BL, Erikson H, Ellsinger B-M, Mikkelson KL,Larsson B. Mortality associated with body fat, fat-free mass andbody mass index among 60-year old Swedish men – a 22-yearfollow-up. Int J Obes Rel Metab Disord 2000; 24: 33–37.38. Lahmann PH, Lissner L, Gullberg B, Berglund G. All-causemortality in relation to body fatness and central adiposity. Int JObes Rel Metab Disord 2000; 24: S25.

Documents

Weight loss causes increased mortality: cons