Table 3. Sensitivity analysis – impact of excluding studies.
| Studies included inmeta-analysis | No. ofstudies | Excluding | Summary estimate(95% CI) | I2-statistic |
| Fatal CVD | ||||
| All studies | 15 | None | 1.79 (1.50 to 2.14) | 71.24% |
| Well-controlled for confounders | 13 | Doğanay, Wannamethee | 1.67 (1.41 to 1.98) | 66.70% |
| Middle-aged or general population studies | 8 | Fernández-Ruiz, Doğanay, Ernsten, Kamphuis, Fang, Tsuji, Pijls | 1.90 (1.51 to 2.39) | 75.41% |
| Elderly population only (65 years and older expect Fang (2003) over 55 years) | 7 | van der Linde, Benjamins, Heidrich, Heistaro, Kaplan, Appels, Wannamethee | 1.50 (1.27 to 1.76) | 0% |
| Male populations only | 9 | van der Linde, Fernández-Ruiz, Doğanay, Benjamins, Fang, Tsuji | 1.74 (1.36 to 2.23) | 70.03% |
| Well-controlled studies based on male populations only | 8 | van der Linde, Fernández-Ruiz, Doğanay, Benjamins, Fang, Tsuji, Wannamethee | 1.52 (1.34 to 1.72) | 0% |
| Both male and female populations | 8 | Pijls, Appels, Kaplan, Heidrich, Kamphuis, Wannamethee | 1.68 (1.32 to 2.12) | 78.56% |
| All except Benjamins | 14 | Benjamins (non-proportional hazards) | 1.73 (1.45 to 2.05) | 51.26% |
| Pre-existing Disease | ||||
| All | 3 | None | 2.42 (1.32 to 4.44). | 71.83% |
| Excluding female studies | 2 | Rutledge | 2.34 (1.09 to 5.06) | 84.95% |
| Non-fatal CVD | ||||
| All studies | 5 | None | 1.66 (0.96 to 2.87) | 83.60% |
| Middle-aged patient population* | 3 | Kennedy, Doğanay | 1.80 (0.81 to 4.02) | 87.14% |
| Well-controlled for confounders | 4 | Doğanay | 1.60 (0.82 to 3.14) | 87.67% |
| All except Van der Linde (2013) | 4 | van der Linde (heavily influences between-study heterogeneity) | 1.34 (0.97 to 1.86) | 27.24% |
| All except Van der Linde (2013) and Doğanay (2012) | 3 | van der Linde, Doğanay | 1.23 (0.85 to 1.77) | 27.47% |
| Fatal and non-fatal CVD | ||||
| All studies | 5 | None | 1.90 (1.26 to 2.87) | 68.61% |
| Well-controlled for confounders | 4 | Doğanay | 1.84 (1.10 to 3.09) | 76.26% |
| Middle-aged patient population* | 3 | Pijls, Doğanay | 2.08 (1.21 to 3.56) | 79.42% |
| Male subjects only | 2 | Kuper, Doğanay, van der Linde | 1.39 (0.92 to 2.08) | 0% |
| All except Van der Linde (2013) | 4 | van der Linde | 1.60 (1.21 to 2.13) | 0% |
*Using standard SRH scale only.