Table 2. Effects of smoking on lung cancer mortality before and after stove improvement in Xuanwei, 1976–1996.

			Before stove improvement		After stove improvement
Smoking variables	n a	Lung cancer deaths	HRpre (95% CI) c	P	HRpost (95% CI) d	P	HRinteraction b	P interaction
Duration of smoking of any type
Duration (per year)	21 686		0.995 (0.987–1.002)	0.17	1.009 (1.002–1.017)	0.02	1.015	0.002
<20	2949	188	1.00 (Ref.)		1.00 (Ref.)		1.00
⩾20–<30	6971	270	0.81 (0.61–1.07)	0.14	1.02 (0.77–1.37)	0.88	1.26	0.23
⩾30–<40	5906	434	0.92 (0.69–1.22)	0.57	1.39 (1.05–1.86)	0.03	1.51	0.02
⩾40	5860	341	1.05 (0.75–1.46)	0.79	1.70 (1.23–2.34)	0.001	1.62	0.01
Ptrend				0.85		0.0008		0.005
Smoking type (pack-years)
Cigarettes (per pack-year)	6394	333	0.997 (0.985–1.009)	0.61	1.015 (1.005–1.025)	0.0004	1.018	0.02
Unfiltered	5637	292	0.990 (0.976–1.004)	0.15	1.013 (1.002–1.025)	0.03	1.024	0.009
Filtered	2018	41	1.014 (0.984–1.045)	0.37	1.019 (0.997–1.042)	0.09	1.005	0.77
Chinese long-stem pipe (per (20 g day–1)*year)	8075	635	0.996 (0.991–1.002)	0.20	1.000 (0.996–1.005)	0.92	1.004	0.25
Water pipe (per (20 g day–1)*year)	2092	112	1.004 (0.987–1.022)	0.61	1.007 (0.987–1.028)	0.48	1.003	0.83

Abbreviations: CI=confidence interval; HR=hazard ratio.

^aReduced data sets included exclusive smokers of each type and non-smokers.

^bHR_interaction=HR_post/HR_pre.

^cAdjusted HR (95% CI) before stove improvement.

^dAdjusted HR (95% CI) after stove improvement. Significant results are shown in bold.