Skip to main content

Table 4 Silicosis and tuberculosis (adjusted and crude GIV), GradePro schema

From: The association between silica exposure, silicosis and tuberculosis: a systematic review and meta-analysis

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 e   Pooled relative risk
(95% CI)
 
All
8
Observational Not
serious
Not
serious
Not serious Not
serious
Strong association, dose response gradient 66,982 3.65
(2.79, 4.78)

HIGH
Low/intermediate
4
Observational Not
serious
Not
serious
Not serious Not
serious
Strong association, dose response gradient 59,091 4.68
(3.22, 6.80)
HIGH
High
4
Observational Not
serious
Not
serious
Not serious No
serious
Strong association,
dose response gradient
7891 3.16
(2.31, 4.32)

HIGH
  1. GIV generic inverse variance, CI confidence interval
  2. a Risk of bias: Not downgraded. We used the Newcastle-Ottawa Scale for cohort and case-control studies. We explored confounding in depth and determined that although confounding was approached differently across studies, the overall risk of bias from the studies did not warrant marking down overall
  3. b Inconsistency: Not downgraded. Overall: I2 reduced from 53% to a moderate 35% following removal of outlier [46]. Low/intermediate TB burden countries: I2 reduced from 49 to 14% following omission of the outlier. High TB burden countries: I2 a moderate 35%, not explained by study design
  4. c Indirectness: Not downgraded. Exact diseases of interest were studied, and the studies covered different industries and country populations
  5. d Imprecision: Not downgraded. Large sample sizes with 95% confidence intervals in the strong effect range
  6. e Other considerations: Large effect and dose response gradient. We marked up for large effect as the associations were strong (relative risk > 2.5 to very strong > 4), although with a wide range. While we cannot exclude the possibility that further studies in different settings might alter the overall effect estimate, the clear dose response in five studies suggests that this is unlikely