From: Methodologies used to estimate tobacco-attributable mortality: a review
Peto's et al method | Prevalence-based analysis in case-control studies | Basic method | |
---|---|---|---|
Variation respect prevalence-based analysis in cohort studies |
Problem: Smoking prevalence is a poor proxy for cumulative hazards of smoking. Solution: Defining SIR (Smoking impact ratio or Synthetic prevalence) authors avoid prevalence limitations. |
Problem: RR extrapolation to different populations than the original is inconsistent. Solution: Designing a case-control study OR could be assessed. |
Problem: RR extrapolation to different populations than the original is inconsistent. Solution: RR can be estimated applying a calculation procedure. |
Calculation procedure |
$SIR=\frac{{C}_{LC}-{N}_{LC}}{{S}_{LC}^{*}-{N}_{LC}^{*}}$
where C_{LC}, N_{LC}, S*_{LC}, N*_{LC} are age-sex specific lung cancer mortality rates for smokers and never smokers in the study and in the reference population (*). |
$OR=\frac{{a}_{1}{b}_{0}}{{a}_{0}{b}_{1}}RR=\frac{{a}_{1}(1-{p}_{1})}{{a}_{0}{p}_{1}}$
where p_{1} is the prevalence between the cases a_{1} is exposed cases b_{1} is exposed controls a_{0} is non exposed cases b_{0} is non exposed controls |
Packs-function in smokers
RRs = 1 + ac((t - 5))-t_{0}) Multistage-function in smokers RRs = 1 + [(t - 5)^{4.5} + ac(1 + 2ac)((t - 5)-t_{0})^{4.5} + 2ac((t - 5)^{4.5}-t_{0} ^{4.5})]/(t - 5)^{4.5} Where a is a constant, c is the number of packs of cigarettes smoked per year, t is the current age and t _{0} is age at start of smoking. In former smokers t_{1} replaces (t - 5) and t_{1} is age at stop smoking. |