Skip to main content

Table 2 Design heterogeneity across 11 cohorts assessing risk of type 2 diabetes, stratified by inclusion of artificially sweetened beverages in the study variable definition

From: Evidence-based mapping of design heterogeneity prior to meta-analysis: a systematic review and evidence synthesis

  Only sugar-sweetened beverages (SSB) (eight cohorts) Soft drinks (SD) (three cohorts)
Study design and population and characteristics N % n %
Study location     
  United States 6 75% 0  
  Europe 1 13% 1 33%
  Japan 1 13% 2 67%
Gender     
  Women 3 38% 0  
  Men 2 25% 0  
  Both men and women 3 38% 3 100
Case of T2D     
  1 to 4,999 2 25% 1 33%
  500 to 4,999 4 50% 2 67%
  5,000+ 2 25% 0  
Duration of follow-up     
  <10 years 4 50% 0  
  10 to 14 years 1 13% 3 100%
  15+ years 3 38% 0  
Mean baseline body mass index (kg/m2)     
  <24 1 13% 2 67%
  24 to 26 4 50% 0  
  >26 3 38% 1 33%
Number/timing of beverage assessment     
  Once at baseline (study length range from 7 to 16 years) 4 50% 3 100%
  Twice (6-year interval) 1 13% 0  
  Every 4 years 3 38% 0  
Proportion of study participants reporting ≥ serving/day     
  10% or fewer or low consumption 2 25% 3 100%
  Between 11 and 15% 3 38% 0  
  More than 15% 2 25% 0  
  Not reported 1 13% 0  
Method of type 2 diabetes (T2D) diagnosis     
  Self report with validation 4 50% 2 67%
  Direct measurement/medical records 4 50% 1 33%
Operation of study variable in multivariable models: highest versus lowest category of consumption
Highest consumption category:     
  2+ drinks or cups/dayARIC,BWHS 2 25% 0 33%
  1+ glasses or servings/day 6 75% 1 67%
  <1 serving/day 0   2  
Lowest consumption category:     
  Never 2 25% 1 33%
  never or rarely 5 63% 2 67%
  <1 cup/dayARIC 1 13% 0  
Characterization of multivariable models Range Median Range Median
No multivariable models presented 1 to 8 4 2 to 4 3
Maximum number of co-variables in multivariable modelsa 9 to 17 14 14 to 17 15
Maximum% change in SSB-T2D risk from age-adjusted estimate 46 to 95% 61% 2 to 26% 18%
  1. aCovariates most frequently adjusted for in multivariable models of the 11 eligible cohorts include physical activity (11 of 11), smoking (11), energy intake (11), BMI (10), family history (9), alcohol intake (8), education (5), and diet quality score (4).