Table 4.
Association between anthropometric changes and incidence rate of diabetes by sex: results from Cox regression*
| Women | Men | Risk Diff. | |||||
| HR † | 95% CI | HR † | 95% CI | RD † | 95% CI | P value | |
| Per absolute changes | |||||||
| ∆Weight | n=12 330 (Power=0.99) | n=7298 (Power=0.89) | n=19 628 (Power=1.00) | ||||
| per SD gain | 1.13 | 1.08 to 1.19 | 1.20 | 1.14 to 1.27 | 0.07 | −0.02 to 0.14 | 0.32 |
| ∆BMI | n=12 324 (Power=0.99) | n=7292 (Power=0.89) | n=19 616 (Power=1.00) | ||||
| per SD gain | 1.14 | 1.08 to 1.19 | 1.23 | 1.15 to 1.31 | 0.08 | −0.03 to 0.17 | 0.10 |
| ∆WC | n=12 311 (Power=0.99) | n=7283 (Power=0.89) | n=19 594 (Power=1.00) | ||||
| per SD gain | 1.18 | 1.12 to 1.24 | 1.25 | 1.18 to 1.34 | 0.07 | −0.02 to 0.15 | 0.26 |
| ∆WHR ‡ | n=12 287 (Power=0.99) | n=7226 (Power=0.88) | n=19 513 (Power=1.00) | ||||
| per SD gain | 1.08 | 1.05 to 1.11 | 1.17 | 1.11 to 1.22 | 0.09 | 0.03 to 0.13 | 0.006 |
| Per relative changes | |||||||
| ∆Weight‡ | n=12 330 (Power=0.99) | n=7298 (Power=0.89) | n=19 628 (Power=1.00) | ||||
| per 5% gain | 1.14 | 1.08 to 1.19 | 1.23 | 1.17 to 1.32 | 0.09 | 0.01 to 0.18 | 0.08 |
| ∆BMI | n=12 324 (Power=0.99) | n=7292 (Power=0.89) | n=19 616 | ||||
| per 5% gain | 1.14 | 1.08 to 1.20 | 1.21 | 1.14 to 1.28 | 0.07 | −0.02 to 0.15 | 0.17 |
| ∆WC | n=12 311 (Power=0.99) | n=7283 (Power=0.89) | n=19 594 (Power=1.00) | ||||
| per 5% gain | 1.20 | 1.14 to 1.27 | 1.28 | 1.19 to 1.37 | 0.08 | −0.02 to 0.16 | 0.26 |
| ∆WHR‡ | n=12 287 (Power=0.99) | n=7226 (Power=0.88) | n=19 513 (Power=1.00) | ||||
| per 5% gain | 1.09 | 1.06 to 1.13 | 1.19 | 1.13 to 1.27 | 0.10 | 0.03 to 0.17 | 0.009 |
| By categories | |||||||
| ∆Weight | n=12 330 (Power=0.99) | n=7298 (Power=0.89) | n=19 628 (Power=1.00) | ||||
| >5% loss | 0.69 | 0.54 to 0.88 | 0.63 | 0.46 to 0.86 | −0.06 | −0.38 to 0.35 | 0.90 |
| ±5% | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. | - |
| >5% gain | 1.34 | 1.09 to 1.64 | 1.54 | 1.25 to 1.91 | 0.20 | −0.13 to 0.54 | 0.50 |
| ∆BMI | n=12 324 (Power=0.99) | n=7292 (Power=0.89) | n=19 616 (Power=1.00) | ||||
| >5% loss | 0.69 | 0.54 to 0.88 | 0.70 | 0.52 to 0.93 | 0.10 | −0.30 to 0.47 | 0.74 |
| ±5% | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. | - |
| >5% gain | 1.30 | 1.07 to 1.60 | 1.58 | 1.29 to 1.95 | 0.28 | −0.08 to 0.61 | 0.18 |
| ∆WC | n=12 311 (Power=0.99) | n=7283 (Power=0.89) | n=19 594 (Power=1.00) | ||||
| >5% loss | 0.67 | 0.52 to 0.87 | 0.52 | 0.39 to 0.71 | −0.15 | −0.47 to 0.15 | 0.24 |
| ±5% | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. | - |
| >5% gain | 1.46 | 1.21 to 1.77 | 1.41 | 1.13 to 1.76 | −0.05 | −0.28 to 0.27 | 0.71 |
| ∆WHR | n=12 287 (Power=0.99) | n=7226 (Power=0.88) | n=19 513 (Power=1.00) | ||||
| >5% loss | 0.72 | 0.54 to 0.96 | 0.53 | 0.40 to 0.71 | −0.18 | −0.50 to 0.09 | 0.20 |
| ±5% | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. | – |
| >5% gain | 1.50 | 1.22 to 1.84 | 1.44 | 1.13 to 1.83 | −0.06 | −0.31 to 0.29 | 0.74 |
*The study power was calculated based on the reported number of participants in each analyses to detect a HR of 1.1 or greater in Cox regression analyses, with the assumption that the overall event rate (diabetes) is 5.0 per 1000 person-year.
†HR and RD were estimated using the Cox regression with anthropometric changes from baseline to either 2004 or 2008 as time-varying exposure (independent) variable and incidence of diabetes as outcome (dependent) variable, after accounting for age, ethnicity, household income, initial anthropometrics, smoking status, physical activity, HEI 2005 Canada tertiles (low/medium/high), parental history of diabetes, and baseline comorbidities (0/1/≥2).
‡The difference between men and women was statistically significant with p<0.01 in Wald’s test of the interaction of the association of anthropometric changes with sex in Cox regression models.
BMI, body mass index; HEI, Healthy Eating Index; RD, risk difference; WC, waist circumference; WHR, waist-hip-ratio.