|
Certainty assessment
|
No. of participants
|
Effect
|
Certainty
|
|---|
Country TB burden
No. of studies
|
Study design
|
Risk of bias a
|
Inconsistency b
|
Indirectness c
|
Imprecision d
|
Other considerations
| |
Pooled relative risk (95% CI)
| |
|---|
All
5
|
Observational
|
Not serious
|
Not serious
|
Not serious
|
Not serious
|
None
|
13,617
|
1.92 (1.36, 2.73)
| ⨁⨁◯◯LOW |
Low/intermediate
2
|
Observational
|
Not serious
|
not serious
|
Not serious
|
Not serious
|
None
|
6879
|
2.75 (1.70, 4.45)
| ⨁⨁◯◯LOW |
High
3
|
Observational
|
Not serious
|
not serious
|
Not serious
|
Not serious
|
None
|
6738
|
1.66 (1.04, 2.66)
| ⨁⨁◯◯LOW |
- GIV generic inverse variance, CI confidence interval
- a Risk of bias: Not downgraded. The overall risk of bias was low across four out of five studies measuring silica exposure. Non-differential exposure misclassification arising from use of proxies for silica exposure is likely to reduce rather than increase any true effect
- b Inconsistency: Not downgraded. Overall I2 reduced from 59% to zero after adjustment for study design by omission of the two case-control studies (one of which had a high risk of bias). Low/ intermediate TB countries): I2 = 0%. High TB countries: I2 was reduced from 73% to zero after adjusting for study design by omission of the one case-control study
- c Indirectness: Not downgraded as exposure metrics were well established proxies of silica exposure, the exact disease of interest was studied, and the studies covered miners in South Africa, foundry workers in Denmark and a variety of silica exposure occupations in Iran
- d Imprecision: Not downgraded. Large sample sizes and 95% confidence intervals exclude the null, i.e. sufficient precision
