Table 3. Summary findings of the meta-analyses for the association between DM and active TB, according to study design.
| Studies | Study population | Effect estimate | Pooled estimate | Heterogeneity measures | |||||
|---|---|---|---|---|---|---|---|---|---|
| Total | Total | Measure of association | Range | Summary estimate | 95% CI | Q (p-value)1 | τ2 2 | I2 3 | |
| Study design | |||||||||
| Prospective | 4 | 872,269 | RRs/HR | 2.24–7.58 | 3.59 | 2.25–5.73 | 13.55 (p = 0.004) | 0.1619 | 77.9% |
| Retrospective | 16 | 56,990,255 | RRs/RR/HR | 1.00–4.23 | 1.55 | 1.39–1.72 | 65.45 (p < 0.0001) | 0.0220 | 77.1% |
| Retrospective | 3 | 9,949 | OR | 1.42–3.59 | 1.69 | 1.35–2.12 | 1.87 (p = 0.392) | 0.0000 | 0.0% |
| Case-control | 17 | 250,720 | OR | 1.16–7.83 | 2.09 | 1.71–2.55 | 77.88 (p < 0.0001) | 0.1010 | 79.5% |
| Cross-sectional | 3 | 327,952 | OR | 1.40–3.17 | 1.70 | 1.28–2.24 | 2.81 (p = 0.245) | 0.0184 | 28.9% |
| Other4 | 1 | 21,230 | RR | 6.00 | 6.00 | 5.00–7.20 | –5 | –5 | –5 |
| Overall6 | 44 | 58,472,375 | OR/RRs/RR/HR | 1.00–7.83 | 2.00 | 1.78–2.24 | 451.95 (p < 0.0001) | 0.0945 | 90.5% |
1 Q: Cochran Q statistic is a measure assessing the existence of heterogeneity in estimates of association between TB and DM.
2 τ2: the estimated between–study variance in the true association between TB and DM estimates. The τ2 is for the variance of beta not for the back-transformed estimate.
3 I2: a measure assessing the magnitude of between-study variation that is due to differences in the association between TB and DM estimates across studies rather than chance.
4 Study by Ponce-de-Leon A., et al.,[64] neither categorized as prospective, retrospective, cross–sectional, or case–control study. Effect estimate is the individual study effect estimate.
5 Meta-analysis was not conducted due to limited number of studies (one study).
6 Overall estimate including risk ratios, rate ratios, hazard ratios, and odds ratios, that is regardless of the measure of association and study design. Background incidence rate of TB did not exceed 2 per 100 person-year in studies estimating an OR, therefore it is reasonable to assume that TB is sufficiently rare so that the ORs would estimate the risk ratios. Pooled estimate was implemented using a random-effects model.
RRs: relative risk; OR: odds ratio; HR: hazard ratio; RR: rate ratio; CI: confidence interval.