Systematic review with meta-analysis of the epidemiological evidence relating smoking to type 2 diabetes

Table 7 Meta-analysis random effects relative risks for ever (vs never) smoking

Grouping1		n2	RR (95%CI)	P
Overall		100	1.25 (1.21-1.28)	P < 0.001, P < 0.01
Sex	Female	24	1.25 (1.18-1.31)
	Male	36	1.25 (0.20-1.31)
	Combined	40	1.22 (1.14-1.31)	P < 0.05
Continent	Asia	41	1.30 (1.25-1.36)
	Europe	36	1.21 (1.17-1.26)
	North and South America	20	1.19 (1.13-1.26)
	Oceania	3	0.87 (0.48-1.57)	P < 0.05
Publication year	Up to 2005	13	1.25 (1.16-1.34)
	2005-2014	47	1.26 (1.20-1.33)
	2015 or later	40	1.23 (1.18-1.28)	NSc
Basis of diagnosis	Self-report only	10	1.35 (1.17-1.56)
	Medical records only	51	1.22 (1.18-1.27)
	Both	39	1.26 (1.19-1.33)	NS
Population	General	95	1.24 (1.21-1.28)
	Pre-diabetics only	1	3.30 (1.24-8.77)
	Pre-diabetics excluded	4	1.43 (1.17-1.76)	P < 0.1
Number of adjustment factors	0	23	1.18 (1.06-1.32)
	1 to 5	16	1.28 (1.20-1.36)
	6 to 10	40	1.24 (1.19-1.30)
	11 or more	21	1.22 (1.16-1.28)	NS
Cohort size	< 5000	39	1.26 (1.14-1.38)
	5000 to 20000	17	1.27 (1.17-1.38)
	> 20000	44	1.24 (1.20-1.28)	NS
Number of type 2 diabetes cases	< 500	46	1.26 (1.16-1.36)
	500 to 999	17	1.32 (1.19-1.47)
	1000 to 2000	9	1.28 (1.14-1.43)
	2001+	28	1.22 (1.17-1.26)	NS
Highest age at baseline	< 60	13	1.35 (1.23-1.47)
	60-74	27	1.32 (1.23-1.41)
	75+	60	1.21 (1.17-1.25)	P < 0.05
Length of follow-up (yr)	< 5	14	1.21 (1.15-1.26)
	5-10	56	1.29 (1.22-1.35)
	> 10	30	1.21 (1.15-1.28)	NS
Definition of smoking	Cigarette	48	1.22 (1.18-1.26)
	Smoking	50	1.28 (1.20-1.36)
	Tobacco	2	1.09 (0.94-1.25)	P < 0.1
Adjusted for age	No	27	1.19 (1.09-1.31)
	Yes	73	1.24 (1.20-1.28)	NS
Adjusted for sex	No	77	1.26 (1.22-1.30)
	Yes	23	1.20 (1.13-1.27)	NS
Adjusted for BMI	No	35	1.23 (1.13-1.34)
	Yes	65	1.24 (1.20-1.28)	NS
Adjusted for physical activity	No	43	1.26 (1.20-1.32)
	Yes	57	1.24 (1.19-1.29)	NS
Adjusted for alcohol consumption	No	46	1.24 (1.19-1.30)
	Yes	54	1.25 (1.20-1.30)	NS
Adjusted for family history of diabetes-	No	62	1.22 (1.17-1.27)
	Yes	38	1.28 (1.23-1.33)	NS
Adjusted for education	No	67	1.29 (1.24-1.34)
	Yes	33	1.17 (1.12-1.21)	P < 0.001
Adjusted for diet	No	76	1.26 (1.22-1.31)
	Yes	24	1.21 (1.15-1.26)	NS
Adjusted for blood pressure	No	57	1.25 (1.19-1.32)
	Yes	43	1.23 (1.19-1.27)	NS
Adjusted for cholesterol	No	75	1.24 (1.20-1.28)
	Yes	25	1.27 (1.19-1.36)	NS
Adjusted for glucose	No	79	1.23 (1.19-1.27)
	Yes	21	1.31 (1.25-1.37)	P < 0.05
Adjusted for triglycerides	No	83	1.23 (1.20-1.27)
	Yes	17	1.31 (1.22-1.41)	NS
Adjusted for waist circumference	No	84	1.25 (1.21-1.30)
	Yes	16	1.21 (1.12-1.31)	NS
Adjusted for other factors	No	42	1.24 (1.15-1.33)
	Yes	58	1.23 (1.20-1.28)	NS

¹For sex, publication year, basis of diagnosis, number of adjustment factors, definition of smoking and age adjusted the grouping relates to characteristics of the relative risk. For other factors it relates to characteristics of the study.

²Number of estimates combined.

³NS means not significant, P ≥ 0.1. For the overall analysis, the first P value relates to heterogeneity between estimates and the second to publication bias. For the other analyses it relates to a test of heterogeneity between the random-effects estimates at each level. Information on publication bias by level of each factor studied is given in Supplementary file 8. CI: Confidence interval; RR: Relative risk; NS: Not significant.