Factors related to increases in body mass index among Canadian adults

1
Factors related to increases in body mass index
among Canadian adults

• Philippe Finès, Statistics Canada
• Heather Orpana, Statistics Canada
• CDPAC, November 6, 2006

2
Purpose

• The purpose of this study is to examine factors
  associated with an important increase of BMI
• in a longitudinal manner
• in an exploratory fashion

3
Past research

• Cigarette smoking, some vigorous activities and
  regular alcohol intake were significantly related
  to BMI decreases over ten years (Kahn et al.
  1997).
• Increased physical activity over 2-3 years
  resulted in smaller weight gains that were
  maintained for 5 years (Schmitz et al., 2000).
• In general, risk factors vary by sex.

4
Object of analysis

• We are interested in the following transition
• BMI increases by at least 2 units
• Why 2 units?
• In the context of an exploratory analysis, we use
  this cutoff value as a measure that represents a
  weight gain that is clinically significant.
• Example Man, 60 kg, 1m 70 (Normal weight BMI
  category)
• Initial BMI20.8
• BMI 2 ? BMI22.8 ? New weight 65.8 kg ?
  Increase of 5.8 kg

5
Design of the analysis (1)

• Survey NPHS
• Cycles 2-6 1996/97, 1998/99, 2000/01, 2002/03,
  2003/2005
• Study population Men and women
• at least 20 years old and at most 56 years old in
  cycle 2,
• with BMI in cycle 2 among these 3 categories
  Normal weight, Overweight, Obese
• Women who were pregnant were excluded for the
  cycles during which they were pregnant.

6
Design of the analysis (2)

• Grouping of population by
• Initial BMI
• Sex
• NB
• At each cycle, BMI was computed as
• Weight / (Initial height)2
• Height and weight were self-reported

7
Model used (1)

• Cox regression
• Data set person-period data
• event (dependent variable)
• 0 (No) if BMI did not increase by at least 2
  units
• 1 (Yes) if it did for the 1st time. In this
  case, all subsequent cycles of this individual
  are ignored.
• Bootstrap weights for an individual were applied
  to each of that individuals records.

8
Model used (2)
Socio-demographic variables (cycle 2) Initial
age (categorical), Initial Income, Initial
Region of residence
Initial BMI (real value)
Socio-demographic variables (cycle i) Marital
status, Education
Health variables (cycle i) Self-perceived
health, Functional limitations
Event (in cycle i1) 0 or 1
Behaviour variables (cycle i) Alcohol
consumption, Smoking, Physical activity in
leisure time, Daily physical activity
9
Sizes of the groups
Sample in the study Total 7,976
Population in the study Total 12,714,938
10
Percentages of persons with BMI2
11
Results Factors associated with BMI2 (1)

• Age (Ref40)
• Age 20-29
• Normal-weight males (RR1.6), Overweight males
  (RR1.6),
• 
  Normal-weight
  females (RR1.3)
• Age 30-39 Obese females (RR1.6)
• Smoking (RefNever)
• Former smoking Overweight males (RR1.5)
• Occasional smoking
• Obese males (RR3.4), Overweight females (RR2.3)
• Daily smoking Normal-weight females (RR1.3)
• Real value of BMI in cycle 2
• Overweight males (RR1.2), Obese males (RR 1.1)
• Normal-weight females (RR 1.1)
• Obese females (RR 0.9 Protective factor)

12
Results Factors associated with BMI2 (2)

• Marital status (RefMarried)
• Separated-divorced-widowed Overweight males
  (RR1.5)
• Income quintile (RefHighest)
• Middle quintile Normal-weight males (RR1.5)
• Highest level of education (RefPost-secondary
  diploma)
• Partial post-secondary level Obese females
  (RR1.7)

13
Results Factors associated with BMI2 (3)

• Self-perceived health (RExcellent/Very good)
• Good level Overweight males (RR1.3)
• Fair/poor level
• Normal-weight females (RR1.6) ,
• Obese males (RR0.3 Protective factor)
• Alcohol consumption (ROccasional)
• Former consumption
• Overweight males (RR1.5),
• Normal-weight females (RR0.7 Protective factor)

14
Conclusion

• Risk varies according to initial category of BMI
  and sex.
• In general, higher BMI within a category, young
  age and smoking (or former smoking) are risk
  factors.
• There is only one factor observed in Overweight
  females the maximum number of factors observed
  is in Overweight males.
• Some factors have opposite effects, depending on
  the group.

15
Discussion (1)

• These results extend those obtained by Le Petit
  on NPHS, cycles 1-5 among overweight persons,
  the probability of becoming obese is
• higher
• for males with higher BMI, younger, who smoke
  regularly and who suffer activity limitations
• for females with higher BMI
• lower
• for males living in the 2 richest income
  quintiles
• for females living in 2 of the richest income
  quintiles, who drink alcohol occasionally, whose
  usual daily physical activity is to sit or walk.
• Our study extends the previous one and has a
  wider scope.

16
Discussion (2)

• Limits
• Small sample
• Self-reported measures
• We assumed that variables in a given cycle have
  an impact on the subsequent cycle.
• New developments
• Trajectory analysis
• BMI other cutoff
• Analyses with nutrition data and more precise
  physical activities.

17
Thank you!

• The authors
• Philippe Finès philippe.fines_at_statcan.ca
• Heather Orpana
• Health Analysis and Measurement Group,
  Statistics Canada

18
Explanatory variables

• Sociodemographic variables (cycle 2)
• Age (categorical) 20-29, 30-39, 40-56 (R)
• (National) Income quintile High (R),
  High-middle, Middle, Low-middle, Low
• Residence region Atlantic, Québec, Ontario (R),
  Prairies, B.-C.
• Sociodemographic variables (time-varying)
• Marital status Single, Sep-Div-Wid, Marrried
  (R)
• Highest education level ltSec., Sec., partial
  post-sec, Post-sec diploma (R)
• Health variables (time-varying)
• Self-perceived health Fair-Poor, Good, Very
  good-Excellent (R)
• Activity limitations Yes, No (R)
• Behaviour variables (time-varying)
• Alcohol consumption Former, Occas. (R),
  Frequent, Regular, Never
• Smoking Former, Occas., Daily, Never (R)
• Physical activity in leisure Reg.Intense (R),
  Reg.Mod., Reg.Light, Occ./Unfreq.
• Daily physical activity Heavy objects (R),
  Light objects, Standing, Sitting