Table 3.
Absolute rates and relative risks of all-cause mortality among Aboriginal current- and past-smokers in the 45 and Up Study, relative to never-smokers, overall and by age group
| Deaths | P-years | Crude rate | RR1 (95% CI) | RR2 (95% CI) | ||
|---|---|---|---|---|---|---|
| Total | 162 | 14 586 | 11.11 | — | — | |
| Overall (summary) | Never-smoker | 41 | 6214 | 6.60 | 1 (ref) | 1 (ref) |
| N = 1388 | Past-smoker | 73 | 5294 | 13.79 | 1.95 (1.32–2.90) | 1.95 (1.32–2.90) |
| Current-smoker | 48 | 3078 | 15.59 | 3.95 (2.56–6.09) | 3.90 (2.52–6.04) | |
| By age group | ||||||
| 45–64 years | Never-smoker | 7 | 4163 | 1.68 | 1 (ref) | 1 (ref) |
| Past-smoker | 21 | 3510 | 5.98 | 3.60 (1.53–8.47) | 3.62 (1.53–8.55) | |
| Current-smoker | 28 | 2606 | 10.75 | 7.08 (3.08–16.28) | 6.92 (2.98–16.04) | |
| 65–74 years | Never-smoker | 10 | 1456 | 6.87 | 1 (ref) | 1 (ref) |
| Past-smoker | 21 | 1294 | 16.23 | 2.49 (1.16–5.33) | 2.53 (1.18–5.45) | |
| Current-smoker | 12 | 381 | 31.50 | 4.86 (2.09–11.32) | 5.51 (2.31–13.15) | |
| ≥75 years | Never-smoker | 24 | 595 | 40.33 | 1 (ref) | 1 (ref) |
| Past-smoker | 31 | 489 | 63.35 | 1.06 (0.58–1.94) | 0.95 (0.50–1.78) | |
| Current-smoker | 8 | 92 | 87.13 | 1.84 (0.80–4.20) | 1.97 (0.83–4.66) | |
The Cox regression models for ‘Overall’ RR1 and RR2 violated the proportional hazard assumption when alpha = 0·05. As age is used for the underlying time scale of these models, the violation indicates an interaction between age and smoking status. Given the observed age differences in the smoking-mortality relationship, and the resulting proportional hazard assumption violation, a sensitivity analysis was undertaken with these models modified to have follow-up time as the underlying time scale, adjusted for 5-year age groups; see Table S3. Results of this sensitivity analysis were consistent with the results from the original ‘Overall’ models with the proportional hazards assumption violation, supporting the robustness of the original models. Given these consistent results, and because the original models allow finer adjustment for age, with age being one of the strongest predictors of mortality, we have retained the original models for the main results despite the proportional hazard violation.
RR1: adjusted for age as the underlying time variable and sex; RR2: additionally adjusted for education and remoteness. Rates are presented per 1000 person-years.
P-years, person-years; RR, relative risk; 95% CI, 95% confidence interval.