Data were analysed for a total of 93 countries among the 4 income groups (low-income=23; lower-middle-income=26; upper-middle-income=23; high-income=21). With regards to overweight and obesity prevalence data, Kiribati showed the highest overweight and obesity prevalence (78.8% and 46.0% respectively), while India showed the lowest overweight prevalence (3.9%) and Bangladesh showed the lowest prevalence for obesity (3.6%). For the period 2014-6 years, the per capita percentage of fat supply ranged from 10.5-41.2% of total caloric supply, whereas the lowest and highest percentages of per capita fat supply were reported for Madagascar and Australia, respectively.
Table 1 summarizes the correlations based on economic status as defined by the GNI for the included 93 countries. A strong positive Pearson'scorrelation coefficient was observed between both the prevalence of overweight (r=0.64, p<0.001) and obesity (r=0.59, p<0.001) with per capita fat supply.When countries were categorized based on income; significant positive correlations were exhibited for both overweight and obesity prevalence (r=0.42, p=0.03 for both) in the lower-middle-income countries whereas, a significant positive correlation was observed only for overweight prevalence (r=0.53, p=0.01) in the high-income countries.
Table 1: Correlation coefficient and coefficient of determination between per capita fat supply and dependent variables of overweight and obesity based on the economic strata classification.
Correlation
|
Overweight
|
Obesity
|
|
r
|
p
|
R2
|
r
|
p
|
R2
|
All countries
|
0.64
|
< 0.001
|
0.41
|
0.59
|
<0.001
|
0.34
|
Income status
|
|
|
|
|
|
|
Low
|
0.23
|
0.28
|
0.05
|
0.29
|
0.17
|
0.09
|
Lower middle
|
0.42
|
0.03
|
0.17
|
0.42
|
0.03
|
0.18
|
Upper middle
|
0.24
|
0.27
|
0.06
|
0.28
|
0.08
|
0.08
|
High
|
0.53
|
0.01
|
0.28
|
0.38
|
0.08
|
0.14
|
r = Pearson’s correlation coefficient, p= Significance, R2=coefficient of determination
Analysis of all countries
The relationship between per capita fat supply and prevalence of both overweight and obesity for all the countries is noted to be logarithmic with strong correlations (Figures 1a and1b respectively). The overweight prevalence of all included countries showed a significant positive correlation (r=0.64, p<0.001) and 41% of the data fit the regression model between per capita fat supply and overweight prevalence (R2=0.41) (Figure 1a).The obesity prevalence of all included countries also showed a significant positive correlation (r=0.59, p<0.001) with the per capita fat supplyand 34% of the data fit the regression model between per capita fat supply and obesity prevalence (R2=0.34) (Figure 1b). The regression lines generated from the correlation analysis showed an upward trend, and as expected, almost all includedcountries scattered around both lines, with very few countriesas outliers (including Egypt and Kiribati) The lower end of these both lines was densely populated by most of the low-income and lower-middle-income countries, except for a few countries as outliers(including Kiribati, Egypt, Algeria, El Salvador, Bolivia). The upper ends of both lines were greatly populated by most of the high-income countries. All upper-middle-income countries (excluding China) scattered around the middle to the upper end of the regression line.
Analysis based on GNI
Scatter plots showing the relationship between the above variables in each income category based on GNI were also produced (Figures 2a-2h). All regression linesgenerated from the correlation analysis also showed an upward trend.
Analysis of low-income countries
Among the low-income countries, Haiti showed the highest prevalence of both overweight (54.9%) and obesity (22.7%) while Ethiopia showed the lowest prevalence of both overweight (20.9%) and obesity (4.5%). However, Gambia and Madagascar showed the highest (27.3%) and lowest (10.5%) per capita fat supply among the low-income group respectively. Both overweight and obesity prevalence were not significantly correlated with per capita fat supply (r=0.23, p=0.28 and r=0.29, p=0.17 correspondingly) (Figure 2a and 2b). However, the correlation effect significantly changed after removing the outliers (Yemen, Haiti) from the analysis, which then gave a significant correlation at both overweight (r=0.49, p=0.02) and obesity (r=0.67, p<0.001) prevalence with 24% and 45%of variations for overweight (R2=0.24) and obesity (R2=0.45) respectively (Supplementary Material: Table 3).
Analysis of lower-middle-income countries
The prevalence of overweight ranged from 19.7% (India) to 78.8% (Kiribati), while obesity ranged from 3.6% (Bangladesh) to 46.0% (Kiribati). A wide range of per capita fat supply, representing 11.3% Bangladesh and 30.3% Kiribati was found in this group. It is noteworthy thatKiribati had the highest prevalence of overweight and obesity, as well as the highest per capita fat supply. In this income category, both overweight and obesity prevalence were significantly correlated with per capita fat supply (r=0.42, p=0.03 and r=0.42, p=0.03 respectively) with the variation of 17% for overweight(R2=0.17) and 18% for obesity (R2=0.18) (Figure 2c and2d).
Analysis of upper-middle-income countries
Fiji had the highest prevalent country of both overweight (30.2%) and obesity (63.8%), while China was the lowest country for those values (32.3% and 6.2% respectively). Per capita fat supply was ranged from 16.2% (Peru) to 35.7% (Belarus) among the upper-middle-income group. Countries in the upper-middle-income group did not show a significant correlation between per capita fat supply and prevalence of both overweight and obesity (r=24, p=0.27 and r=0.28, p=0.08 respectively) (Figure 2e and 2f). However, after removing one outlier (China), significant correlation reported with obesity prevalence (r=0.43, p=0.04) with 18% of variation (R2=0.18) (Supplementary Material: Table 3). All upper-middle-income countries expect a couple of countries such as Georgia and China were clustered close to the regression line.
Analysis of high-income countries
Among high-income countries, Japan showed the lowest prevalence for both overweight (27.2%) and obesity (4.3%). New Zealand presented the highest overweight prevalence (65.6%) whereas the Bahamas showed the highest obesity prevalence (31.6%). The per capita fat supply ranged from 27.1% (Chile) to 41.6% (Australia). Only overweight prevalence significantly correlated with per capita fat supply (r=0.53, p=0.01)with 28% of variation (R2=0.28) (Figure 2g). However, per capita fat supply did not significantly correlate with the obesity prevalence (r=0.38, p=0.08) in the high-income group (Figure 2h). Almost all included countries in the regression line generated from the correlation analysis among the high-income group scattered around the line, with two countries, (Republic of Korea and Japan) as outliers.