Table 5.
Multivariable meta-regression model of ASMR estimates from premature CVD mortality
| Covariates | Adjusted β (95% CI) | t value | P value |
|---|---|---|---|
| CVD types | |||
| Stroke | Ref. | ||
| IHD | 1.55 (0.73, 2.37) | 3.80 | < 0.001 |
| Other heart disease | 0.92 (0.16, 1.68) | 2.44 | 0.019 |
| Total CVD | 1.67 (0.77, 2.58) | 3.73 | < 0.001 |
| Sex | |||
| Female | Ref. | ||
| Male | 1.01 (0.57, 1.43) | 4.70 | < 0.001 |
| Country income classificatione | |||
| HIC | Ref. | ||
| MIC | 1.20 (0.10, 2.30) | 2.19 | 0.033 |
| Age group | |||
| 0 – 70 years | Ref. | ||
| 20 – 64 years | 1.41 (0.86, 1.96) | 5.15 | < 0.001 |
| 30 – 74 years | 2.14 (1.22, 3.06) | 4.69 | < 0.001 |
Meta-regression model was applied using a backward method for variable selection. The Hartung-Knapp adjustment, which accounts for the heterogeneity among studies, was applied. Adjusted covariates included CVD types, sex, country income classification, and age group. No multicollinearity or interaction was present. Model fitness was checked based on the corrected Akaike Information Criterion (AICc). The reduced model (AICc = 161.38) showed better fit compared to the full model, which included study time (AICc = 166.72)