Paternal Smoking and Risk of Childhood Acute Lymphoblastic Leukemia: Systematic Review and Meta-Analysis

Table 2

Results of meta-analysis of paternal smoking in different time periods and childhood acute lymphoblastic leukemia (ALL) risk.

Paternal smoking^a	Studies included	𝑁	Fixed-effect model^b		Random-effect model		𝐼^2  ^c
			OR; (95% CIs)	𝑃	OR; (95% CIs)	𝑃	(%)
Paternal ever smoking	[21, 23, 26–38, 41–43]	18	1.11; (1.05, 1.16)	.000	1.11; (1.05, 1.18)	.000	18
Overall lifetime ever smoking^d	[21, 23, 28, 36, 38, 42]	6	1.07; (1.01, 1.14)	.027	—	—	0
Preconception	[21, 23, 27–30, 32–34, 37, 41–43]	13	1.16; (1.03, 1.31)	.016	1.25; (1.08, 1.46)	.002	53
With the highest exposure index	[27–30, 33, 34, 37, 41–43]	10	1.37; (1.13, 1.66)	.001	1.38; (1.11, 1.72)	.004	45
Removing the smallest and greatest ORs	[21, 23, 27, 28, 30, 32–34, 41–43]	11	1.16; (1.03, 1.31)	.013	1.25; (1.08, 1.46)	.003	50
Removing the two greatest ORs	[21, 23, 27–30, 32–34, 37, 41–43]	11	1.15; (1.03, 1.28)	.013	1.20; (1.05, 1.36)	.003	41
Removing the highest weight^e	[21, 23, 27–30, 33, 34, 37, 41–43]	12	1.25; (1.08, 1.45)	.003	1.31; (1.10, 1.56)	.003	49
Removing the OR from Rudant, 2008^f	[21, 23, 27–30, 32–34, 37, 41–43]	12	1.10; (1.00, 1.22)	.060	1.15; (1.01, 1.31)	.03	26
With well-defined exposure period^g	[21, 23, 28, 30, 32, 34, 37, 41–43]	10	1.11; (0.99, 1.23)	.069	1.17; (1.02, 1.35)	.026	33
With paternal smoking during 1 or 3 months before pregnancy	[21, 23, 30, 34, 43]	5	1.13; (0.98, 1.29)	.085	—	—	0
With population-based controls	[21, 23, 27, 28, 30, 32–34, 41, 43]	10	1.16; (1.02, 1.31)	.020	1.25; (1.07, 1.45)	.005	54
During pregnancy	[21, 27–29, 33–35, 42]	8	1.19; (1.07,1.32)	.001	1.24; (1.07,1.43)	.004	54
With the highest exposure index	[27–29, 33]	4	1.34; (1.02, 1.77)	.037	1.28; (0.93, 1.76)	.13	65
With adjusted ORs	[21, 27, 28, 33, 34, 42]	6	1.26; (1.04, 1.51)	.017	1.34; (1.07, 1.68)	.010	60
Removing the smallest and greatest ORs	[21, 27, 28, 33–35]	6	1.19; (1.06, 1.33)	.002	1.25; (1.07, 1.45)	.003	58
Removing the highest weight^h	[21, 27–29, 33, 34, 42]	7	1.23; (1.03,1.47)	.022	1.28; (1.04,1.58)	.022	59
Removing the OR from Rudant, 2008	[21, 27–29, 34, 35, 42]	7	1.15; (1.06, 1.23)	.000	—	—	0
With well-defined exposure period^i	[21, 34, 35, 42]	4	1.15; (1.07, 1.25)	.002	1.16; (1.03, 1.31)	.015	25
With population-based controls	[21, 27, 28, 33–35]	6	1.19; (1.06, 1.33)	.002	1.25; (1.07, 1.45)	.003	58
After birth	[21, 26, 30, 31, 33, 41, 42]	7	1.20; (0.97, 1.49)	.092	1.24; (0.96, 1.60)	.092	64
With the highest exposure index	[26, 30, 31, 33, 41, 42]	6	1.35; (1.06, 1.72)	.008	1.33; (1.00, 1.78)	.05	57
Removing the smallest and greatest ORs	[26, 30, 31, 33, 41]	5	1.32; (1.03, 1.70)	.027	1.27; (0.95, 1.68)	.10	58
Removing the highest weight^j	[26, 30, 31, 33, 41, 42]	6	1.35; (1.06, 1.72)	.008	1.33; (1.00, 1.78)	.05	57
Removing the OR from Rudant, 2008	[21, 26, 30, 31, 41, 42]	5	1.05; (0.89, 1.23)	.58	1.06; (0.89, 1.26)	.25	12
With population-based controls	[21, 30, 31, 33, 41]	5	1.23; (0.96, 1.59)	.11	1.25; (0.92, 1.69)	.15	68

^aWhen multiple indices of exposure categories were available, the highest was selected to estimate the summary effects; otherwise, the binary category was selected, except for the analysis of the subgroup with highest exposure;
^b95% confidence interval (CI) and 𝑃 values were estimated by Shore correction when they were wider or greater than the unadjusted estimates by fixed-effect models. The Shore correction incorporates interstudy heterogeneity;
^c𝐼²=100%×(𝑄−𝑑𝑓)/𝑄, where Q is Cochran’s heterogeneity statistic and df the degrees of freedom, with negative values of 𝐼² put equal to zero. 𝐼² describes the percentage of total variation across studies that is due to heterogeneity rather than chance
^dOnly included studies that reported an association between childhood ALL and paternal overall smoking status during lifetime
^eRemoved the OR from Pang et al. 2003, which accounted for 42% of the weight;
^fRudant et al. 2008 reported risk of childhood ALL for paternal smoking one year before the child birth to the time of interview; the same estimated risk was used for calculating the summary effect of childhood ALL for paternal smoking before conception, during pregnancy and after birth.
^gOnly included ORs for paternal smoking before pregnancy, removed the ORs from Rudant et al. 2008, John et al. 1991 and Magnani et al. 1990, which estimated ORs for paternal smoking in the year (12 months) before birth, preconception, and during the prenatal period;
^hRemoved the OR from Sorahan et al. 1995, which accounted for 70% of the weight;
ⁱOnly included ORs for paternal smoking during pregnancy and excluded ORs for paternal smoking during the year (or 12 months) prior to birth.
^jRemoved the OR from Brondum et al. 1999, which accounted for 35% of the weight.