A REVIEW OF THE SCIENCE UNDERLYING STUDIES OF OBESITY

1
A REVIEW OF THE SCIENCE UNDERLYING STUDIES OF
OBESITY

• By
• Sencer Ecer
• and
• Richard Higgins
• LECG,LLC
• 1725 Eye Street, NW, Suite 800
• Washington, DC 20006
• E-Mail Rhiggins_at_Lecg.com

2
WHO AND THE BUSH ADMINISTRATION

• WHO Report of April 03
• Recommendations
• HHS Letter
• Quality and objectivity of science are inadequate
• Policy recommendations not justified by
  underlying science

3
CONTENTS

• The Definition of Obesity
• Proximate Causes of Obesity
• Causes of Overeating and an Unbalanced Diet
• Consequences of Obesity

4
THE DEFINITION OF OBESITY

• In Terms of Body Fat or BMI
• BMI Weight in LBS X .704.5/Height in
  inches-squared
• You Know It When You See It
• Top Percentiles
• BMI that Minimizes the Risk of
• Mortality?
• Morbidity?
• The role of multivariate analysis
• Style of life
• Key Reference Robert Fogel, AER, 1994.

5
Economic Growth, Population Theory, and
Physiology The Bearing of Long-Term Processes on
the Making of Economic Policy

• By
• Robert Fogel
• American Economic Review, 1994

6
WAALER CURVES

• Mortality/BMI Waaler curves for Norwegian men and
• women between the ages of 50 and 54 as of 1963
• show that a BMI of 23 maximizes longevity
• Relative risk of mortality is related to BMI
• U-shaped with minima at 23
• The minima vary marginally by age category

7
FOGEL ISO-MORTALITY CURVES

• Combinations of weight and height are associated
• with the same relative risk of mortality
• Can be defined in terms of morbidity too
• Obesity is typically defined as BMI gt 30, where
  the ideal, 30, is associated with best
  performance measured in terms of mortality (or
  morbidity)
• Optimal BMI is derived from the weight and height
  that minimizes mortality (or morbidity), here,
  based on the incidence of death in Waalers
  sample of 1.7 million Norwegian adult males (ages
  50 to 64)
• optimal BMI for 6 male is 24 (mortality), 26
  (morbidity)
• optimal BMI for 56 male is 25.7 (mortality),
  26.7 (morbidity)

8
PROXIMATE CAUSES OF OBESITY

• Overeating
• Unbalanced Diet
• energy-dense, added sugar
• excessive salt
• Inadequate Exercise
• Reference Cutler et al, J. Economic
  Perspectives, 2003.

9
Why Have Americans Become More Obese?

• By
• D. Cutler, E. Glaeser and J. Shapiro
• Journal of Economic Perspectives
• Summer 2003

10
FINDINGS (I)

• Daily distribution of food intake by occasion
  Continuing Survey of Food Intake, 1977-78 and
  1994-1996
• Comparisons
• calorie intake for men rose by 268 per day for
  women, 143 per day
• most of the increases are explained by increased
  calorie consumption through snacks 241 for men,
  160 for women

11
FINDINGS (I) (CONT.)

• Implications
• Obesity does not result from increased portion
  sizes in restaurants
• Consumption of fattening foods at fast food
  restaurants has not caused increased obesity

12
FINDINGS (II)

• Energy expenditure from time-use data (Robinson
• and Godbey, Time for Life, Penn State University
• Press, 1997).
• Data Time-use diaries self reporting.
• Overall daily activity time has not declined
  between 1965 and 1995
• Time used in recreation increased over this
  period from 27 minutes/day to 47 minutes/day

13
FINDINGS (II) (cont.)

• Hypotheses
• Increased obesity results from greater food
  consumption, not less exercise
• Principal cause is a reduction in the time-cost
  of food, which has led to food consumption on
  more occasions
• Several Implications are tested (but with very
  limited data)
• Cross-country regression percent of adult pop
  that is obese as a function of (1) the price of a
  Big Mac (2) National income per capita and (3)
  Percent of females in the labor force. Price of
  Big Mac is negative and significant.

14
FINDINGS (II) (cont.)

• Authors also suggest that processing technology
  has not only
• succeeded in the lowering the full price of food
  which
• encourages consumption but that it has reduced
  the time
• between the decision to consume and the act of
  consumption.
• They hypothesize that many individuals lack self
  control (they
• are not responsible individuals) they are less
  willing to trade
• near-term consumption for somewhat more remote
• consumption when the choice is imminent than when
  it is
• remote.
• How prevalent is such Irrationality? Is it
  present in children more than in adults?

15
CAUSES OF OVEREATING AND AN UNBALANCED DIET

• Reduced Time-Cost of Food
• Technology
• Government subsidies
• Media Influence
• Advertising
• Ignorance
• Self-Control Problems

16
ADVERTISING AND OBESITY REVIEWS AND REVIEWS OF
REVIEWS
17
KAISER FAMILY FOUNDATION

• The Role of Media in Childhood Obesity (2004)
• Conclusions
• Majority of research supports the proposition
  that more media time is positively related to
  being overweight
• How media contribute to childhood obesity is
  inconclusive
• Nonetheless, it appears likely that food
  advertising may well be the principal
  contributor

18
STRATHCLYDE REVIEW (2003)

• Conclusions
• Advertising to children has an adverse effect on
  food preferences, purchasing behavior and
  consumption

19
AMERICAN ACADEMY OF PEDIATRICS

• Policy Statement Children, Adolescents, and
  Advertising, Pediatrics (1995) Conclusions
• Advertising directed toward children is
  inherently deceptive

20
A REVIEW OF THE STRATHCLYDE REVIEW

• David Ashton, Food Advertising and Childhood
• Obesity, Journal of the Royal Society of
  Medicine,
• Feb. 2004 Conclusions
• The two most highly touted studies in the
  Strathclyde Review do not support the alleged
  positive relation between TV advertising or food
  promotion and food consumption behavior of
  children
• There is no good evidence that advertising has
  substantial influence on childrens food
  consumption

21
A REVIEW OF THE STRATHCLYDE REVIEW (cont.)

• Although food advertising to children has been
  banned in Quebec since 1980, childhood obesity
  prevalence does not differ across Canadian
  provinces

22
CONCLUSIONS

• The magnitude of the effects of food advertising
  on
• the demand for food is not known at this time.
  The
• problem is made difficult for two reasons
• The profitability of advertising depends
  principally on the competitive effects of
  advertising, not its effects on total demand.
  (Does advertising principally sell food brands or
  types of food?)
• It is difficult to measure the total, as opposed
  to marginal effects of advertising in the absence
  of good natural experiments (the effects of bans,
  certain interventions, and so on)

23
ADVERTISING AND OVERWEIGHT IN CHILDREN

• Effect of Television Advertisements for Foods on
  Food Consumption in Children by Jason, et al,
  Appetite (2003)
• 42 children (aged 9-11) were divided into Lean,
  Overweight and Obese based on BMI (BMI gt 30,
  obese BMI lt 25, lean), they were shown cartoons
  with and without food advertising on two
  occasions two weeks apart after each showing
  they were presented with four types of snack
  foods and allowed to eat as much as desired.

24
ANALYSIS AND CONCLUSIONS

• Obese and overweight children recognized
  significantly more food ads than the lean
  children
• The heavier the child the more food was eaten
  after seeing the food ads, and the heavier the
  child the more food was eaten after seeing the
  non-food ads
• correlation coefficient between food ads
  recognized and food consumed was 0.47 (not a
  partial correlation coefficient)
• the correlation coefficient between non-food ads
  recognized and food consumed NOT reported

25
ANALYSIS AND CONCLUSIONS

• Regardless of weight the average food intake with
  food ad was significantly greater than that w/o
  food ad
• Why not regression?
• The reason why obese and overweight children
  selectively recognized more food adverts and also
  ate the most food requires further
  investigation.
• The observed association between remembering
  food ads and eating more indicates that a
  susceptibility to food cues contributes to this
  overeating and promotes weight gain in children.

26
Television Viewing as a Cause of Increasing
Obesity Among Children in the United States,
1986-1990

• By
• Steven Gortmaker, Aviva Must, Arthur Sobol,
• Karen Peterson, Graham Colditz and William Dietz,
• Archives of Pediatrics Adolescent Medicine, 1996

27
DATA

• Baseline (1986) and follow-up (1990) interviews
  of 746 children 6 to 11 years old in 1986 with
  complete height and weight data from a nationally
  representative sample of youth (National
  Longitudinal Survey of Labor Market Experience,
  Youth Cohort (NLSY)
• TV viewing time in 1990 0-2 hrs/day, (2-3
  hrs/day, (3-4 hrs/day, (4-5 hrs/day and (5-24
  hrs/day
• Overweight, 1986, 1990 BMI gt 85th percentile
  adjusted for age and gender (using national
  standards)
• Demographic and socioeconomic data

28
MODEL

• Logistics model log odds overweight-not
  overweight regressed on (1) TV viewing time, (2)
  child overweight in 1986, (3) race, (4) mothers
  education, (5) gender, (6) intelligence of child,
  mother, and so on
• Also looked at odds of being overweight in 1990
  if not overweight in 1986 related to similar
  variables, including TV viewing

29
RESULTS

• Odds of being overweight are 5.3 times higher for
  children with more than 5 hours TV viewing a day
  than for those with at most 2 hours of TV viewing
  per day odds ratio of 5.3 and difference is
  statistically significant.
• Becoming overweight at least 5 hrs./day vs. at
  most 2 hrs/day Odds ratio is 8.3, and is
  statistically significant.

30
COMMENTS

• Duh!
• Authors are not clear as to whether childs BMI
  in 1986 is a regressor, or whether mothers BMI
  is a regressor in general, more transparency
  would be helpful.
• difficult to determine how robust findings are
• difficult to evaluate methodology
• Endogeneity obesity causes TV watching
• bias
• authors claim to test for this by looking at how
  TV watching in 1990 relates to overweight in
  1986 they find no statistically significant
  relation
• inadequate test for causality (Granger, Hausman,
  and others)

31
CONCLUSIONS

• TV viewing is only a proximate cause
• TV viewing (studying?, computer game play?)
  substitutes for more energy-intense activities
• increased snacking while watching TV ?
• TV advertising increases the demand for food
  consumption ?
• Concern that there is a cumulative effect of
  weight gain that will increase prevalence of
  chronic diseases adipocytes created by
  overeating ratchet effect
• Solution TVs rigged to generators powered by
  riding bicycles

32
CONSEQUENCES OF BEING OVERWEIGHT

• Effects on Earnings
• Effects on Disability
• Increased Morbidity and Mortality
• Social costs
• Reference J. Cawley, J. Human Resources, 2004

33
The Impact of Obesity on Wages

• By
• John Cawley,
• Forthcoming in the Journal of Human Resources
  (2004)

34
DATA

• NLSY longitudinal survey evidence with annual
• interviews, 1979-1994, and bi-annual interviews,
• 1996-2002

35
STATISTICAL MODEL

• Log Wages a b BMI c X (demographic and
  socioeconomic variables) e
• Estimated separately for Black, White and
  Hispanic women and men

36
MEASUREMENT AND ESTIMATION ISSUES

• When e is correlated with BMI, the regression
  estimate of b is biased
• Wages determine BMI (simultaneity bias)
• Unobservable genetic and non-genetic factors that
  are correlated with BMI and with wage potential
  (errors in variables bias)

37
PREVIOUS FIXES

• Lagged BMI (removes contemporaneous
  interdependence, only)
• In absence of simultaneity, differencing using
  twins data
• Instrumental Variables
• Cawley deals with each, including removal of
  individual fixed effects to deal with
  time-invariant correlation between unobservable
  characteristics that affect both weight and wages

38
RESULTS

• Cannot reject lack of simultaneity bias (Hausman
  test) OLS estimates best
• There is a negative relation between wages and
  weight only for white females (an increase in
  weight of 64 pounds from the average of 148 to
  212, reduces wages by 9, the equivalent of 1.5
  years of education)

39
Body-Mass Index and Mortality in a Prospective
Cohort of U.S. Adults

• By
• E.E. Calle, M.J. Thun, J. Petrelli,C. Rodriguez,
  C. Heath,
• The New England Joural of Medicine, 1999.

40
DATA

• Participants in the Cancer Prevention Study II
  (that was begun by the American Cancer Society in
  1982).
• Over 450,000 men and over 580,000 women were
  included (ages at least 30 years old, with at
  least one household member over 45.

41
STATISTICAL MODEL

• Relative risks of death for four subgroups
• Smokers (if ever smoked) v. non-smokers
• People with history of disease v. no history of
  disease
• History of disease cancer, heart disease,
  stroke, respiratory disease, current illness or a
  weight loss of at least 10 pounds in the previous
  year, current illness (of any type).
• Controls for age, education, other demographics,
• exercise level,
• aspirin use,
• index of fat consumption, vegetable consumption
• (for women) estrogen replacement therapy

42
MEASUREMENT AND ESTIMATION ISSUES

• History of disease variable includes any type of
  current illness. Vague?
• Signs/magnitudes of coefficients are not
  reported. Expected directions? No statistical
  specification tests reported. Correct
  specification?

43
FINDINGS

• Obesity is most strongly associated with
  increased death rate among nonsmokers and those
  with history of disease (Stronger in whites than
  blacks, weakest in black women).
• High BMI most predictive of death from
  cardiovascular disease, especially in men.

44
AUTHORS OWN CRITICISMS

• Antecedent disease bias pre-existing,
  unrecognized disease processes causing death.
• Unable to control for long-term weight loss,
  which might increase relative risk
• Self-reported index overestimates height,
  underestimates weight, thus overestimates BMI
• No direct measure of central adiposity. Where
  fat is located in the body matters in relative
  risk of death. E.g., waist hip ratio predicts
  death better than BMI in Iowa elderly women

45
CONCLUSIONS

• Risk of death from all causes increases
  throughout the range of overweight for all.
• Risk associated with high BMI is higher for
  whites than blacks.