Table 2.
Strength of the evidence for the relationship between chewing tobacco and the seven health outcomes analyzed
| Health outcome | RR (95% UI without γ) | RR (95% UI with γ) | BPRF | ROS | Star rating | Pub. bias | No. of studies | Selected bias covariates | Pair in GBD |
|---|---|---|---|---|---|---|---|---|---|
| Stroke | 1.46 (1.28, 1.68) | 1.46 (1.11, 1.93) | 1.16 | 0.07 |  |
No | 3 | None | N |
| Esophageal cancer | 2.14 (1.77, 2.57) | 2.14 (0.89, 5.15) | 1.02 | 0.01 |  |
No | 22 | Maximally adjusted; Adjusted for smoking, age, and sex | Y |
| Lip and oral cavity cancer | 3.64 (3.00, 4.41) | 3.64 (0.66, 19.95) | 0.87 | -0.07 |  |
No | 70 | Chewing tobacco product; Study subpopulation | Y |
| Larynx cancer | 2.66 (1.98, 3.57) | 2.66 (0.52, 13.63) | 0.68 | -0.20 |  |
No | 24 | Aggregate outcome definition; Adjusted for age and sex | N |
| Nasopharynx cancer | 2.50 (1.79, 3.49) | 2.50 (0.49, 12.66) | 0.64 | -0.22 |  |
No | 17 | Maximally adjusted; Adjusted for age and sex | N |
| Other pharynx cancer | 2.33 (1.80, 3.01) | 2.33 (0.45, 12.04) | 0.59 | -0.27 |  |
No | 31 | Aggregate outcome definition; Adjusted for age and sex | N |
| Ischemic heart disease | 1.30 (0.88, 1.92) | 1.30 (0.29, 5.83) | N/A | N/A | No | 8 | None | N |
The reported relative risk (RR) and its 95% uncertainty interval (UI) reflect the risk an individual who uses chewing tobacco has of developing the outcome of interest relative to that of someone who does not use chewing tobacco. Gamma (γ) quantifies the estimated between-study heterogeneity of included observations. We report two separate 95% UIs, one that is estimated without incorporating between-study heterogeneity (γ) and one that does account for this source of uncertainty—"95% UI with γ.” The Burden of Proof Risk Function (BPRF) is calculated for risk-outcome pairs that were found to have significant relationships at an 0.05 level of significance when between-study heterogeneity is not incorporated. The BPRF corresponds to the 5th quantile estimate of relative risk accounting for between-study heterogeneity closest to the null for each risk–outcome pair, and it reflects the most conservative estimate of excess risk associated with chewing tobacco that is consistent with the available data. Since we define chewing tobacco exposure as a dichotomous risk factor, i.e., an individual either currently chews tobacco or does not, the risk-outcome score (ROS) is calculated as the signed value of natural log(BPRF) divided by two. Negative ROSs indicate that the evidence of the association is very weak and inconsistent. For ease of interpretation, we have transformed the ROS and BPRF into a star rating (1–5) with a higher rating representing a larger effect with stronger evidence. A zero-star rating is assigned to risk-outcome pairs whose RR 95% uncertainty interval without consideration of between-study heterogeneity crosses 1. The potential existence of publication bias, which, if present, would affect the validity of the results, was tested using Egger’s Regression. Included studies represent all available relevant data identified through our systematic reviews from January 1970 through January 2023. The selected bias covariates were chosen for inclusion in the model using an algorithm that systematically detects bias covariates that correspond to significant sources of bias in the observations included. If selected, the observations were adjusted to better reflect the gold standard values of the covariate. See the Supplementary Information for more information about the candidate bias covariates.