The Interaction of Smoking with Gene Polymorphisms on Four Digestive Cancers: A Systematic Review and Meta-Analysis.

The main purpose of this study was to perform a meta-analysis to assess the interaction between smoking and nine genes (GSTM1, GSTT1, GSTP1, CYP1A1, NAT2, SULT1A1, hOGG1, XRCC1 and p53) on colorectal cancer, gastric cancer, liver cancer and oesophageal cancer. Published articles from the PubMed, ISI and EMBASE databases were retrieved. A total of 67 case-control studies or nested case-control studies were identified for the analysis. The pooled jodds ratio (OR) with 95% confidence interval (CI) was calculated using the random effect model. The overall study showed that the GSTM1 polymorphism was associated with the risk of the four digestive cancers among Asian population (OR 1.284, 95% CI: 1.122-1.470, p: 0). Subgroup analyses by cancer site showed that GSTM1 null genotype increased the gastric cancer risk in total population (OR 1.335, 95% CI: 1.145-1.556, p: 0). However, the association of GSTM1 null genotype with the oesophageal cancer risk was found in smokers (OR 1.382, 95% CI: 1.009-1.894, p:0.044), but not in non-smokers (OR 1.250, 95% CI: 0.826-1.891, p:0.290). Moreover, smokers with the CYP1A1 IIe462Val polymorphism were at an increased cancer risk in Asian population (OR=1.585, 95% CI 1.029-2.442, p: 0.037). None of the other gene-smoking interactions was observed in the above cancers. This meta-analysis reveals two potential gene-smoking interactions, one is between smoking and GSTM1 on oesophageal cancer, and the other is between smoking and CYP1A1 IIe462Val on the four cancers in Asian population. Future studies need to be conducted to verify the conclusions.

Introduction

Cancer was the second leading cause of non-communicable diseases deaths worldwide in 2015. Most cancer patients die from digestive cancers between 2005 and 2015, of which the death toll increased to 832,000 for colorectal cancer (CRC), 818.9,000 for gastric cancer (GC), 810.5,000 for liver cancer (LC) and 439,000 for oesophageal cancer (OC)1. Moreover, the incidence of these four cancers ranks in the top ten over the world, mainly in developing countries 2. Especially, these cancers are generally recognized as tobacco-related cancers (TRCs) by the International Association of Research in Cancer (IARC) 3. However, not all individuals exposed to tobacco develop these cancers. Because the etiology of cancer is multifactorial and complicated 4, cigarette smoking, as a prevalent environment factor, may interact with multiple genetic factors, leading to a higher susceptibility to cancer.

The research on the gene-smoking interaction in cancer risk has been popular 5. Previously published studies clarified the molecular mechanism of the gene-smoking interaction. Most tobacco carcinogens first form DNA adducts via metabolic activation; persistent DNA adducts induce mutations in some critical genes and initiate carcinogenesis 6. The elimination of DNA adducts requires DNA repair, implying that variations of the DNA repair genes may be related to different repair efficiencies of DNA damage 7. Moreover, various detoxification pathways are competitive and different individuals have distinct balances between metabolic activation and detoxification, influencing the cancer risk 8. Increasing epidemiologic studies and meta-analyses have indicated the interaction between smoking and gene polymorphisms in various cancer types 9-11. However, most meta-analyses only assessed the interaction between single gene polymorphism and smoking on one or several cancers. Furthermore, the results were inconsistent or even conflicting. Hence, we performed a comprehensive meta-analysis on the interaction of smoking with ten gene polymorphisms in four digestive cancers. The aim was to develop a more powerful evaluation of gene-smoking interaction on major digestive cancers risk.

Materials and methods

Search strategy

PubMed, ISI and EMBASE databases were searched until Dec. 2017 with combinations of the following keywords: “smoke, cigarette, tobacco, smoking”, “gene, polymorphism”, “colorectal, colon, rectum, colorectum, liver, hepatocellular, oesophageal, oesophagus, gastric, stomach”, and “cancer, carcinoma, adenomas”. No restrictions were placed on language. References of the retrieved and review articles were also screened by hand.

Inclusion and exclusion criteria

Studies that were included in our analysis had to meet all of the following criteria: (1) evaluated the gene-smoking interaction on the risk of digestive cancers; (2) only case-control studies or cohort studies were considered; (3) provided case and control or cohort size by gene-smoking interaction; (4) showed the gene polymorphisms that were evaluated in at least five independent studies on the four digestive cancers and (5) when an author had several studies on the same patient population, only the most recent or largest sample article was included..The following exclusion criteria were used: (1) the full text was not obtained; (2) only case population; and (3) duplicated study.

Data extraction and quality assessment

All data were independently extracted by two investigators according to the above selection criteria. The information collected from each study are as follows: the first author's last name; year of publication; country of origin; ethnicity; study design; total number of cases and controls or cohort; cancer type; gene names; number of cases and controls or cohort by gene polymorphisms; number of cases and controls or cohort by gene-smoking interaction. Smoking habits were categorized as non-smoker and smoker. The number of cases and controls or cohort by gene-smoking interaction was extracted according to four combinations: non-smoker + “no risk” polymorphism; non-smoker + “at risk” polymorphism; smoker + “no risk” polymorphism; and smoker + “at risk” polymorphism. For each gene polymorphism, the “at risk” phenotype was identified based on known biological mechanisms and the classification conducted by most included articles. “At risk” polymorphism for GSTM1/GSTT1 was the null (-/-); for GSTP1, the IIe105Val substitution (Ile/Val+Val/Val); for CYP1A1, the 3801T>C substitution (MspI) (T/C+C/C) and Ile462Val substitution (Ile/Val+Val/Val), for NAT2, the fast + intermediate (at least one *4 or *12) acetylator; for SULT1A1, the slow+intermediate (at least one *2) sulphation, for hOGG1, the Ser326Cys substitution (Ser/Cys+Cys/Cys); for XRCC1, the Arg399Gln substitution (Arg/Gln+Gln/Gln); and for p53, the Arg72Pro substitution (Arg/Pro +Pro/Pro).

The quality of each study was evaluated by the Newcastle-Ottawa Scale (NOS), which is a 9-star system containing the following three dimensions: selection; comparability; and outcome (cohort studies) or exposure (case-control studies) 12. A study with 7-9 scores was classified as a high-quality study, while those with scores of 4-6 and 0-3 are moderate- and low-quality studies, respectively 13.

Statistical methods

The reference group was identified as “no risk” polymorphism, and the odds ratios (OR) with 95% confidence intervals (CI) were calculated to determine a risk of the association between gene polymorphisms and digestive cancers. To be conservative, the random effects model was applied to calculate the summary risk. In addition, the subgroup analyses were conducted based on the cancer site and ethnicity. Heterogeneity was evaluated among studies by calculating the Q-statistic and I value 14. Publication bias was assessed by constructing the funnel plots (there was no publication bias if the funnel plot was symmetric) and quantified using Begg's test and Egger's test 15, 16, in which a p-value<0.05 indicated the presence of potential publication bias. All statistical analyses were performed using Comprehensive Meta-Analysis Software, version v. 2.0 (CMA, Biostat, Englewood, NJ, USA). For the positive findings, the false-positive report probability and statistical power were calculated by G*Power software 17, 18.

Results

Literature search

A total of 1979 articles were collected from the 3 databases. As shown in Figure 1, 1491 publications were excluded; 1251 articles were titles, abstracts, systematic reviews, meta-analyses, case reports and irrelevant articles and another 240 papers lacked data on gene-smoking interactions. Finally, a total of 67 studies were included in this meta-analysis. The reason for removing 421 studies from the remaining articles was that they evaluated the gene polymorphisms in less than five independent studies on the four digestive cancers.

Figure 1

Flow diagram of study selection in this meta-analysis. This flowchart indicates that the process of screening relevant studies based on the inclusion/exclusion criteria. A total of 67 studies were included in this meta-analysis.

Study characteristics and quality assessment

Study characteristics are summarized in Table 1. These studies were case-control or nested case-control studies, including 21,954 cases and 30,341 controls. Forty-three studies were performed in Asia, 11 studies were performed in Europe, 10 studies were performed in the Americas, and 3 studies were performed in Africa. Among all identified articles, 30 evaluated GSTM1 polymorphism 19-48, 18 evaluated GSTT1 polymorphism 20-24, 30-32, 34, 35, 40, 42-48, 12 evaluated GSTP1 polymorphism 11, 22, 30, 32, 34, 35, 42, 49-53, 8 evaluated CYP1A1 IIe462Val polymorphism 9, 27, 28, 54-58, 7 evaluated CYP1A1 MspI polymorphism 26, 28, 45, 54, 57, 58, 8 evaluated NAT2 polymorphism 24, 28, 36, 38, 46, 59-61, 6 evaluated SULT1A1 polymorphism 24, 45, 62-65, 8 evaluated hOGG1 polymorphism 66-73, 7 evaluated XRCC1 polymorphism 52, 67, 69, 74-77, and 6 evaluated p53 polymorphism 78-83.

Table 1

Characteristics of included case-control studies

First author, year	NOS	Country/Ethnicity	Cancer site	Genes	Genotype distribution (cases/controls)		Genotype distribution by smoking status (cases/controls)
					No risk*	At risk$	Non-smoker		Smoker
							No risk*	At risk$	No risk*	At risk$
Wang,2004	7	China/Asia	Oesophagus	GSTM1	53/57^	74/44	24/37	33/26	29/20	41/18
Rudolph,2012	8	German/Europe	Colorectum	GSTM1	822/844	932/923	368/424	425/466	404/382	458/417
				GSTT1	1,433/1,459	313/308	644/722	142/170	715/672	147/123
Lilla,2007	9	Germany/Europe	Colorectum	SULT1A1	212/263	292/340	106/132	132/157	106/131	160/183
Gao,2002	9	China/Asia	Oesophagus	GSTM1	35/90	106/133	13/38	36/58	22/52	70/75
				GSTT1	67/104	74/119	20/44	29/52	47/60	45/67
			Stomach	GSTM1	63/90	90/133	10/38	20/58	53/52	70/75
				GSTT1	82/104	71/119	20/44	10/52	62/60	61/67
Dandara,2006	7	South Africa/Africa	Oesophagus	SULT1A1	115/132	121/134	27/41	28/47	88/91	93/87
Li,2010	7	South Africa/Africa	Oesophagus	GSTM1	206/200	133/80	55/75	8/20	151/125	125/60
				GSTT1	127/178	113/102	27/66	36/29	100/112	77/73
				GSTP1	92/107	148/173	21/30	42/65	71/77	106/108
Gertig,1998	7	America/Americas	Colorectum	GSTM1	97/104	114/117	36/40	41/40	61/64	73/77
				GSTT1	173/169	36/51	61/60	16/19	112/109	20/32
Tiemersma,2004	7	Netherlands/Europe	Colorectum	GSTM1	203/206	228/226	81/102	85/118	119/103	143/108
				GSTT1	370/363	61/69	139/177	27/43	228/185	34/26
				NAT2	262/254	169/178	89/132	66/67	169/121	79/77
				SULT1A1	149/169	282/263	66/97	72/97	83/62	128/106
Abo-Hashem,2016	7	Egypt/Africa	Liver	GSTP1	23/31	17/9	11/27	9/3	12/4	8/6
Li,2005	6	China/Asia	Stomach	GSTM1	33/36	67/26	16/23	30/19	17/13	37/7
Tsukino,2004	7	Japan/Asia	Stomach	hOGG1	32/74	110/197	11/38	39/99	21/36	71/98
Inoue,2000	7	Japan/Asia	Colorectum	GSTM1	97/97	108/123	19/37	17/36	78/60	91/87
				CYP1A1&	86/87	119/133	14/20	22/53	72/67	97/80
Lee,2000	7	China/Asia	Oesophagus	GSTP1	65/160	25/94	11/98	11/50	54/55	14/40
Shen,2005	7	China/Asia	Stomach	GSTM1	41/314	71/361	31/302	54/345	10/12	17/16
				CYP1A1#	70/412	42/264	57/391	29/254	13/21	13/10
Yoshida,2007	7	Japan/Asia	Colorectum	GSTM1	30/59	36/62	20/26	15/29	8/29	18/32
				CYP1A1#	34/79	32/42	20/36	15/19	14/40	12/21
				CYP1A1&	20/49	46/72	8/24	27/31	12/23	14/38
				NAT2	2/9	64/112	0/5	35/50	1/4	25/57
Zendehdel,2009	9	Sweden/Europe	Oesophagus	GSTM1,	52/230	43/239	17/112	13/87	35/127	30/143
				GSTT1	80/394	15/76	24/173	6/26	56/221	9/49
				GSTP1	44/208	50/245	13/82	16/110	31/126	34/135
			Oesophagus	GSTM1,	35/230	42/239	4/112	4/87	30/127	38/143
				GSTT1	70/394	7/76	8/173	1/26	62/221	6/49
				GSTP1	26/208	52/245	5/82	5/110	21/126	47/135
			Stomach	GSTM1,	54/230	70/239	6/112	8/87	4/127	62/143
				GSTT1	111/394	13/76	12/173	2/26	99/221	11/49
				GSTP1	47/208	75/245	6/82	8/110	41/126	67/135
Lee,2006	7	Chile/Americas	Stomach	GSTM1,	60/207	13/56	29/128	2/33	31/79	11/23
				CYP1A1&	38/153	35/110	16/90	15/71	22/63	20/39
Huang,2006	9	America/Americas	Colon	GSTM1	297/503	257/371	111/211	97/151	184/292	158/219
				GSTT1	428/603	130/271	162/247	46/115	259/356	83/155
Moore,2005	7	U.S./Americas	Colorectum	GSTM1	311/313	352/376	105/122	115/150	190/173	217/205
				GSTT1	561/584	129/118	182/230	44/50	350/325	77/56
				GSTP1	282/317	399/381	97/132	123/140	173/171	251/251
Cai,2001	8	China/Asia	Stomach	GSTM1	35/51	60/43	12/28	22/32	23/23	38/11
Tamer,2005	7	Turkey/Asia	Stomach	GSTM1	30/116	40/88	17/75	19/45	13/41	21/43
				GSTT1	49/151	21/53	25/85	11/35	24/66	10/18
				GSTP1	38/90	32/114	20/49	16/71	18/41	16/43
Slattery,2002	7	USA/Americas	Colon	GSTM1	761/892	816/1012	332/413	326/486	429/479	490/526
				NAT2	920/1154	688/804	366/540	298/380	554/614	390/424
García-González,2012	8	Spain/Europe	Stomach	GSTM1	274/290	283/267	125/151	120/147	51/40	71/35
				GSTT1	437/440	120/117	188/228	57/70	97/56	25/19
				GSTP1	255/251	302/306	119/138	126/160	50/36	72/39
Malik,2010	8	India/Aisa	Stomach	hOGG1	50/94	58/101	15/68	17/79	35/21	40/17
Malik,2009	7	India/Asia	Stomach	GSTM1	44/116	64/79	12/85	20/62	32/26	43/12
Slattery,2003	9	US/Americas	Rectum	GSTM1	230/279	243/295	84/123	88/124	145/156	153/171
				NAT2	247/306	204/255	90/143	74/105	156/163	128/150
Kasahara,2008	7	Japan/Asia	Colorectum	hOGG1	17/39	51/82	8/14	28/41	7/23	19/37
				XRCC1	42/62	26/59	20/29	16/26	18/30	8/30
Wu,2003	6	China/Asia	Oesophagus	SULT1A1	135/274	52/34	44/153	25/19	91/121	27/15
Yu,1995	7	China/Asia	Liver	GSTM1	14/55	16/95	7/34	10/61	7/21	6/34
Yu,1999 a	7	China/ Asia	Liver	GSTM1	38/151	42/177	25/94	22/104	13/57	20/73
Yu,1999 b	7	China/Asia	Liver	GSTM1	42/159	42/216	26/91	23/132	16/67	19/84
				GSTT1	42/194	41/181	25/110	24/113	17/84	17/67
Moslehi,2006	6	USA/Americas	Colorectum	NAT2	413/376	272/317	140/158	92/124	249/195	168/168
Malakar,2012	9	India/Asia	Stomach	GSTM1	45/107	57/97	7/52	14/30	38/55	43/67
				GSTT1	65/111	37/93	11/45	10/37	54/66	27/56
Yu,2000	7	China/Asia	Liver	NAT2	27/55	124/156	16/30	59/100	11/25	65/56
Songserm,2014	8	Thailand/Asia	Liver	hOGG1	34/95	111/234	14/55	50/123	20/40	61/111
				XRCC1	4/21	156/318	2/11	70/170	2/10	86/148
Ates,2005	7	Turkey/Asia	Colorectum	GSTM1	83/116	98/88	44/75	46/45	39/41	52/43
				GSTT1	118/151	63/53	56/85	34/35	62/66	29/18
				GSTP1	73/90	108/114	30/49	60/71	43/41	47/43
Van der Hel,2003 a	8	Netherlands/Europe	Colorectum	GSTM1	124/396	88/369	73/271	65/257	51/125	23/112
				GSTT1	154/541	58/224	104/385	34/143	50/156	24/81
Van der Hel,2003 b	7	Netherlands/Europe	Colorectum	NAT2	146/495	112/362	99/341	63/249	42/153	45/113
Moaven,2010	7	Iran/Asia	Oesophagus	GSTP1	84/74	64/62	51/65	50/46	33/10	14/16
Zhang,2014	8	China/Asia	Colorectum	hOGG1	44/48	203/252	30/32	129/158	14/16	74/94
Ghosh,2016	9	India/Asia	Stomach	GSTP1	41/61	29/21	10/38	9/16	31/23	20/5
				XRCC1	28/48	42/34	8/33	11/21	20/15	31/13
Boccia,2005	7	Italy/Europe	Stomach	SULT1A1	40/160	36/100	33/126	24/83	7/31	10/15
Boccia,2015	7	Italy/Europe	Liver	GSTM1	96/139	105/150	31/91	48/81	62/48	57/69
				GSTT1	141/220	60/69	59/129	20/43	81/91	38/26
				CYP1A1&	165/226	56/64	65/136	20/37	98/90	35/27
				SULT1A1	132/180	89/110	52/103	33/70	78/77	55/40
Yuan,2012	8	China/Asia	Liver	hOGG1	67/144	283/256	30/48	83/84	37/96	200/172
Sakamoto,2006	7	Japan/Asia	Liver	hOGG1	56/73	153/202	35/56	105/152	21/17	48/50
Hanaoka,2001	8	Brazil/Americas	Stomach	hOGG1	133/123	75/82	72/85	48/55	61/38	27/25
Gelatti,2005	8	Italy/Europe	Liver	GSTM1	101/185	99/215	41/60	34/80	60/125	65/135
				GSTT1	168/328	32/72	67/124	8/16	101/204	24/56
				NAT2	105/201	95/199	40/65	35/75	65/136	60/124
Setiawan,2000	9	China/Asia	Stomach	GSTM1	45/207	42/212	26/131	18/143	19/76	24/69
				GSTT1	37/228	44/190	18/146	21/127	19/82	23/63
Setiawan,2001	9	China/Asia	Stomach	GSTP1	61/296	20/123	30/199	10/75	31/97	10/48
Zhang,2012	7	China/Asia	Stomach	GSTP1	331/343	219/207	69/136	37/77	82/100	59/71
Chen,2004	8	China/Asia	Colon	GSTM1	23/151	30/188	17/92	16/108	6/57	14/79
				GSTT1	41/270	12/69	26/153	7/47	15/116	5/20
			Rectum	GSTM1	33/151	39/188	23/92	26/108	10/57	13/79
				GSTT1	61/270	11/69	43/153	6/47	18/116	5/20
Bhat, 2014	7	India/Asia	Oesophagus	CYP1A1#	253/300	273/226	101/134	99/122	152/166	174/104
Chen, 2011	7	China/Asia	Stomach	XRCC1	177/132	157/202	83/88	67/124	94/44	90/78
Fernandes, 2016	8	Brazil/Americas	Colorectum	CYP1A1#	193/312	34/88	107/190	24/53	86/122	10/35
				CYP1A1&	165/246	62/154	96/156	35/87	69/90	27/67
Hou, 2005	7	USA/Americas	Colorectum	CYP1A1#	633/643	42/36	219/258	9/19	387/344	29/15
Li, 2009	7	China/Asia	Liver	CYP1A1#	560/598	410/402	313/320	223/212	247/278	187/190
Little, 2006	8	Northeast Scotland/Europe	Colorectum	CYP1A1#	235/372	16/24	75/128	5/5	84/142	7/10
				CYP1A1&	190/310	42/68	63/107	12/19	68/122	16/27
Malakar,2014	7	India/Asia	Stomach	p53	11/36	94/174	1/14	20/71	10/22	74/103
Qiu, 2016	6	China/Asia	Liver	p53	221/244	764/748	137/207	488/645	84/37	276/103
Shao, 2008	6	China/Asia	Oesophagus	p53	163/195	510/499	61/90	229/219	102/105	281/280
Shen,2004	7	China/Asia	Stomach	p53	96/94	228/223	36/46	97/76	60/48	131/147
Yan,2009	6	China/Asia	Stomach	XRCC1	241/345	214/305	121/186	91/163	106/155	111/136
Yang, 2008	7	China/Asia	Oesophagus	p53	373/273	62/277	222/200	43/200	151/73	19/77
Yu, 1999 c	9	China/Asia	Liver	CYP1A1#	46/239	35/170	33/147	15/97	13/92	20/73
				CYP1A1&	25/152	56/257	19/86	29/158	6/66	27/99
Yu,2004	7	China/Asia	Oesophagus	XRCC1	65/88	70/64	33/50	28/35	32/38	42/29
Cai, 2017	7	China/Asia	Liver	p53	63/65	279/282	33/55	146/171	30/10	133/111
Putthanachote,2017	7	Putthanachote/ Asia	Stomach	XRCC1	12/8	89/194	8/3	41/105	4/5	48/89

Aberrations: NOS, the Newcastle-Ottawa-Scale.

^Number of cases and controls.

*The wild type of each gene.

$The mutant type of each gene.

# For CYP1A1, the IIe462Val substitution (IIe/Val+Val/Val).

& For CYP1A1, the 3801T>C substitution (MspI) (T/C+C/C).

As shown in Table 1, the quality scores of studies ranged from 6 to 9. Therefore, 91% of the studies (n=61) were high-quality studies (studies with a score≥7).

Tobacco metabolizing related genes

GST genes

Among 30 studies on the GSTM1 polymorphism in Table 2, the results showed the GSTM1 null genotype increased the four digestive cancers risk (OR=1.118, 95% CI 1.022-1.222). No significant publication bias was found using Begg's test (p=0.10), while there was publication bias by Egger's test (p=0.045). According to the trim and fill analysis, the adjusted estimated effect was OR 1.054 (95% CI: 0.954-1.163) based on the random-effects model. Substantial heterogeneity was observed in this analysis (Q=70.248, p=0.000, I2= 53.024 %), which suggested that GSTM1 polymorphisms have different effects on the risk of four cancers, depending on the cancer type and ethnicity. Subgroup analysis based on ethnicity revealed that such an association was observed among both African (OR=1.614, 95% CI 1.038-2.51; I2=0%, p for heterogeneity=1) and Asian (OR=1.284, 95% CI 1.122-1.47; I2=57.181%, p for heterogeneity=0.001) populations; further subgroup analysis based on the cancer type showed that the GSTM1 null genotype were associated with an increased risk of oesophageal cancer (OR=1.406, 95% CI 1.124-1.759; I2=63.644%, p for heterogeneity=0.027) and gastric cancer (OR=1.335, 95% CI 1.145-1.556; I2=52.921%, p for heterogeneity=0.019). Stratified analysis by smoking status showed the association of the GSTM1 null genotype with the four cancers risk was significant among smokers (OR=1.179, 95% CI 1.030-1.349; I2=57.328%, p for heterogeneity=0). In subgroup analyses among smokers, there was publication bias (p Begg =0.004; p Egger =0.029). According to the trim and fill analysis, the adjusted estimated effect was OR 1.012 (95%CI: 0.867-1.181) based on the random-effects model. However, the effect size was only found in Asian population (OR=1.355, 95% CI 1.089-1.686; I2=39.566%, p for heterogeneity=0.044). Smokers with the GSTM1 null genotype had an increased risk of oesophageal cancer (OR=1.382, 95% CI 1.009-1.894, I2=55.082, p for heterogeneity=0.064) and gastric cancer (OR=1.690, 95% CI 1.298-2.201, I2=69.955%, p for heterogeneity=0). Moreover, subgroup analyses in non-smokers showed that the GSTM1 null genotype also increased the gastric cancer risk (OR=1.344, 95% CI 1.054-1.715; I2=51.576%, p for heterogeneity=0.024). The GSTM1 null genotype was associated with the four cancers risk in Asian population (OR=1.237, 95% CI 1.020-1.500; I2=44.307%, p for heterogeneity=0.023), no publication bias was observed (p>0.05).

Table 2

Meta-analysis of the association between GSTM1, GSTT1 polymorphisms and the four digestive cancers risk

Stratified analysis	Subgroup analysis	No. of studies	OR (95% CI)	Heterogeneity test			Publication bias	False-positive report probability	Statistical power
				Q	P	I2 (%)	p
GSTM1 total population	Overall cancer	30	1.118(1.022-1.222)	70.248	0	53.024	0.100*	0.050	0.659
	Cancer type						0.045$
	Colorectum	12	1.010(0.911-1.121)	11.808	0.461	0
	Oesophagus	4	1.406(1.124-1.759)	11.002	0.027	63.644		0.047	0.337
	Stomach	11	1.335(1.145-1.556)	21.241	0.019	52.921		0.048	0.991
	Liver	5	0.866(0.691-1.086)	1.763	0.779	0
	Ethnicity
	Africa	1	1.614(1.038-2.51)	0	1	0		0.042	0.782
	Americas	6	1(0.853-1.172)	3.552	0.616	0
	Asia	17	1.284(1.122-1.47)	39.702	0.001	57.181		0.048	0.976
	Europe	7	0.991(0.862-1.141)	7.724	0.461	0
GSTM1 non-smokers	Overall cancer	30	1.071(0.948-1.210)	54.333	0.011	39.263	0.486*
	Cancer type						0.186$
	Colorectum	12	0.993(0.847-1.163)	10.507	0.572	0
	Oesophagus	4	1.250(0.826-1.891)	6.091	0.192	34.331
	Stomach	11	1.344(1.054-1.715)	20.651	0.024	51.576		0.047	0.716
	Liver	5	0.866(0.622-1.206)	8.996	0.061	55.538
	Ethnicity
	Africa	1	0.545(0.207-1.435)	0	1	0
	Americas	6	0.956(0.759-1.205)	7.012	0.220	28.698
	Asia	17	1.237(1.020-1.500)	30.524	0.023	44.307		0.048	0.542
	Europe	7	1.018(0.828-1.253)	8.301	0.405	3.625
GSTM1 smokers	Overall cancer	30	1.179(1.030-1.349)	77.335	0	57.328	0.004*	0.050	0.728
	Cancer type						0.029$
	Colorectum	12	1.014(0.855-1.203)	12.204	0.429	1.673
	Oesophagus	4	1.382(1.009-1.894)	8.905	0.064	55.082		0.046	0.301
	Stomach	11	1.690(1.298-2.201)	33.284	0	69.955		0.047	0.999
	Liver	5	0.862(0.606-1.227)	3.146	0.534	0
	Ethnicity
	Africa	1	1.725(0.891-3.339)	0	1	0
	Americas	6	1.035(0.794-1.349)	1.146	0.950	0
	Asia	17	1.355(1.089-1.686)	28.106	0.044	39.566		0.048	0.755
	Europe	7	1.054(0.826-1.343)	37.431	0	78.628
GSTT1 total population	Overall cancer	18	0.970(0.863-1.092)	38.800	0.010	45.876	0.150*
	Cancer type						0.628$
	Colorectum	8	0.935(0.782-1.119)	17.558	0.025	54.438
	Oesophagus	3	1.068(0.778-1.466)	7.426	0.060	59.599
	Stomach	6	0.923(0.722-1.180)	8.715	0.121	42.626
	Liver	3	1.084(0.772-1.521)	2.108	0.348	5.136
	Ethnicity
	Africa	1	1.553(0.978-2.465)	0	1	0
	Americas	3	0.827(0.643-1.063)	8.377	0.015	76.124
	Asia	8	1.017(0.837-1.237)	11.380	0.181	29.703
	Europe	8	0.950(0.805-1.122)	8.447	0.391	5.297
GSTT1 non-smokers	Overall cancer	18	0.979(0.838-1.143)	28.943	0.115	27.443	0.554*
	Cancer type						0.610$
	Colorectum	8	0.881(0.752-1.031)	9.792	0.280	18.297
	Oesophagus	3	1.845(1.204-2.829)	4.065	0.255	26.196		0.043	0.999
	Stomach	6	0.973(0.732-1.293)	4.639	0.462	0
	Liver	3	0.965(0.649-1.436)	0.047	0.977	0
	Ethnicity
	Africa	1	3.034(1.564-5.889)	0	1	0		0.040	0.909
	Americas	3	0.797(0.605-1.051)	3.882	0.144	48.487
	Asia	8	0.999(0.779-1.280)	9.695	0.287	17.481
	Europe	8	0.944(0.801-1.112)	1.915	0.984	0
GSTT1 smokers	Overall cancer	18	0.977(0.843-1.132)	31.747	0.062	33.852	0.888*
	Cancer type						0.996$
	Colorectum	8	1.043(0.834-1.305)	13.100	0.108	38.930
	Oesophagus	3	0.858(0.593-1.240)	4.475	0.215	32.963
	Stomach	6	0.844(0.615-1.159)	8.526	0.130	41.354
	Liver	3	1.192(0.778-1.825)	2.556	0.279	21.741
	Ethnicity
	Africa	1	1.181(0.638-2.186)	0	1	0
	Americas	3	0.864(0.606-1.232)	6.401	0.041	68.754
	Asia	8	1.117(0.844-1.478)	10.680	0.221	25.093
	Europe	8	0.907(0.710-1.158)	12.237	0.141	34.627

The bold letters show statistically significant results.

* Begg's test for publication bias.

$ Egger's test for publication bias.

Among 18 studies on the GSTT1 polymorphism in Table 2, we found that the GSTT1 null genotype could increase the oesophageal cancer risk in non-smokers (OR=1.845, 95% CI 1.204-2.829; I2=26.196%, p for heterogeneity=0.255). By subgroup analysis in non-smokers, Only one study showed the GSTT1 polymorphisms were related to the risk of four cancers in African population (OR=3.034, 95% CI 1.564-5.889)22. No publication bias was detected in this analysis (p>0.05).

Among 12 studies on the GSTP1 polymorphism in Supplementary Table S1, no significant correlations were found except one study on liver cancer in non-smokers (OR=7.364, 95% CI 1.671-32.440)49. There was no publication bias (p>0.05).

CYP1A1 gene

Eight papers provided data on the CYP1A1 IIe462Val polymorphism in Table 3. The results indicated that smokers with the CYP1A1 Ile462Val polymorphisms were at an increased risk of four cancers in Asian population (OR=1.585, 95%CI 1.029-2.442; I2=41.870%, p for heterogeneity=0.142). Seven articles were about CYP1A1 MspI polymorphism in Supplementary Table S1. The CYP1A1 MspI polymorphisms were not associated with the risk of four cancers in stratified analysis and subgroup analysis.

Table 3

Meta-analysis of the association between CYP1A1, SULT1A1 polymorphisms and the four digestive cancers risk

Stratified analysis	Subgroup analysis	No. of studies	OR (95% CI)	Heterogeneity test			False-positivereport probability	Statisticalpower
				Q	P	I2 (%)
CYP1A1 IIe462Val total population	Overall cancer	8	1.102(0.911-1.332)	13.969	0.052	49.888
	Cancer type
	Colorectum	4	1.039(0.724-1.490)	8.418	0.038	64.360
	Oesophagus	1	1.432(0.829-2.474)	0	1	0
	Stomach	1	0.936(0.494-1.776)	0	1	0
	Liver	2	1.082(0.714-1.639)	0.005	0.945	0
	Ethnicity
	Americas	2	0.851(0.572-1.267)	3.949	0.047	74.676
	Asia	5	1.197(0.961-1.492)	6.270	0.180	36.201
	Europe	1	1.055(0.505-2.207)	0	1	0
CYP1A1 IIe462Val non-smokers	Overall cancer	8	0.973(0.827-1.145)	6.539	0.478	0
	Cancer type
	Colorectum	4	0.901(0.586-1.384)	3.519	0.318	14.759
	Oesophagus	1	1.077(0.647-1.792)	0	1	0
	Stomach	1	0.783(0.434-1.412)	0	1	0
	Liver	2	0.964(0.665-1.397)	1.531	0.216	34.681
	Ethnicity
	Americas	2	0.720(0.460-1.127)	0.539	0.463	0
	Asia	5	1.009(0.846-1.204)	3.349	0.501	0
	Europe	1	1.707(0.478-6.089)	0	1	0
	Overall cancer	8	1.341(0.959-1.876)	17.436	0.015	59.853
	Cancer type
CYP1A1 IIe462Valsmokers	Colorectum	4	1.067(0.565-2.015)	9.278	0.026	67.667
	Oesophagus	1	1.827(0.658-5.075)	0	1	0
	Stomach	1	2.100(0.494-8.926)	0	1	0
	Liver	2	1.385(0.636-3.014)	1.846	0.174	45.832
	Ethnicity
	Americas	2	0.883(0.431-1.807)	8.182	0.004	87.778
	Asia	5	1.585(1.029-2.442)	6.881	0.142	41.870	0.046	0.932
	Europe	1	1.183(0.341-4.108)	0	1	0
SULT1A1 total population	Overall cancer	6	1.315(1.009-1.715)	17.371	0.004	71.216	0.048	0.993
	Cancer type
	Colorectum	2	1.137(0.649-1.994)	0.502	0.478	0
	Oesophagus	2	1.724(0.940-3.163)	13.122	0	92.379
	Stomach	1	1.440(0.579-3.579)	0	1	0
	Liver	1	1.103(0.480-2.536)	0	1	0
	Ethnicity
	Africa	1	1.036(0.730-1.472)	0	1	0
	Asia	1	3.104(1.923-5.011)	0	1	0	0.044	0.997
	Europe	4	1.148(0.984-1.339)	1.332	0.722	0
SULT1A1non-smokers	Overall cancer	6	1.257(0.849-1.861)	17.039	0.004	70.656
	Cancer type
	Colorectum	2	1.068(0.494-2.311)	0.021	0.885	0
	Oesophagus	2	2.027(0.853-4.819)	10.933	0.001	90.853
	Stomach	1	1.104(0.339-3.591)	0	1	0
	Liver	1	0.934(0.296-2.947)	0	1	0
	Ethnicity
	Africa	1	0.905(0.461-1.776)	0	1	0
	Asia	1	4.575(2.308-9.070)	0	1	0	0.041	0.991
	Europe	4	1.045(0.836-1.307)	0.242	0.970	0
SULT1A1smokers	Overall cancer	6	1.248(0.952-1.637)	8.766	0.119	42.964
	Cancer type
	Colorectum	2	0.996(0.660-1.501)	0.439	0.507	0
	Oesophagus	2	1.454(0.893-2.369)	3.562	0.059	71.922
	Stomach	1	2.952(0.864-10.091)	0	1	0
	Liver	1	1.357(0.688-2.680)	0	1	0
	Ethnicity
	Africa	1	1.105(0.652-1.875)	0	1	0
	Asia	1	2.393(1.117-5.126)	0	1	0	0.040	0.694
	Europe	4	1.146(0.854-1.539)	4.358	0.225	31.169

The bold letters show statistically significant results.

SULT1A1 gene

In Table 3, the SULT1A1 slow/intermediate phenotypes were associated with a 31.5% increase in the risk of four cancers (OR=1.315, 95% CI 1.009-1.715) from 6 studies. However, such an association was not observed in stratified analysis and subgroup analysis. Only one paper showed the association was significant in Asian population (OR=3.104, 95% CI 1.923-5.011)64.

NAT2 gene

Eight papers provided data on the NAT2 polymorphism, as shown in Table 4. Two studies indicated that the NAT2 polymorphism was associated with the risk of four cancers in Asian population (OR=1.701, 95% CI 1.019-2.838) 28, 60. Moreover, the association was also observed in smokers (OR=2.513, 95% CI 1.156-5.462).

Table 4

Meta-analysis of the association between NAT2 polymorphism and the four digestive cancers risk

Stratified analysis	Subgroup analysis	No. of studies	OR (95% CI)	Heterogeneity test			False-positive report probability	Statistical power
				Q	P	I2 (%)
total population	Overall cancer	8	0.990(0.872-1.125)	11.662	0.112	39.978
	Cancer type
	Colorectum	6	0.970(0.837-1.123)	7.981	0.157	37.351
	Liver	2	1.115(0.796-1.561)	3.280	0.070	69.515
	Ethnicity
	Americas	3	0.961(0.832-1.109)	6.101	0.047	67.219
	Asia	2	1.701(1.019-2.838)	0.303	0.582	0	0.044	0.426
	Europe	3	0.963(0.793-1.168)	0.553	0.759	0
non-smokers	Overall cancer	8	1.047(0.889-1.232)	8.934	0.257	21.649
	Cancer type
	Colorectum	6	1.072(0.894-1.285)	7.483	0.187	33.179
	Liver	2	0.886(0.556-1.414)	0.695	0.404	0
	Ethnicity
	Americas	3	1.044(0.802-1.359)	2.478	0.290	19.295
	Asia	2	1.251(0.597-2.622)	1.607	0.205	37.779
	Europe	3	1.004(0.730-1.382)	4.444	0.108	54.996
smokers	Overall cancer	8	0.993(0.817-1.205)	13.717	0.056	48.968
	Cancer type
	Colorectum	6	0.933(0.755-1.152)	7.364	0.195	32.101
	Liver	2	1.334(0.837-2.125)	4.334	0.037	76.927
	Ethnicity
	Americas	3	0.913(0.749-1.113)	2.470	0.291	19.037
	Asia	2	2.513(1.156-5.462)	0.113	0.737	0	0.040	0.379
	Europe	3	0.988(0.745-1.310)	4.613	0.100	56.648

The bold letters show statistically significant results.

DNA repair genes

Neither hOGG1 gene nor XRCC1 gene polymorphism was not associated with the risk of four cancers, as shown in Supplementary Table S1.

Tumour suppressor gene

We also found no significant association of p53 polymorphism with the risk of four cancers (Supplementary Table S1).

Discussion

A total of 67 case-control studies on the interaction of gene-smoking on the risk of four digestive cancers were identified in this review. This study included six tobacco metabolizing genes (GSTM1, GSTT1, GSTP1, CYP1A1, SULT1A1, and NAT2), two DNA repair genes (hOGG1 and XRCC1) and one tumour suppressor gene (p53). To the best of our knowledge, this is the first meta-analysis that investigated the joint effect of the most gene polymorphisms and smoking on four digestive cancers. Our data indicated the GSTM1 polymorphism was associated with the risk of four digestive cancers among Asian population (OR 1.284, 95% CI: 1.122-1.470). The GSTM1 null genotype could increase the gastric cancer risk (OR 1.335, 95% CI: 1.145-1.556) in total population. However, the association of the GSTM1 null genotype with the oesophageal cancer risk was found in smokers (OR 1.382, 95% CI: 1.009-1.894), not in non-smokers (OR 1.250, 95% CI: 0.826-1.891). Interestingly, we found the GSTT1 null genotype could increase the oesophageal cancer risk among non-smokers in only 3 studies (OR 1.845, 95% CI: 1.204-2.829). The SULT1A1 polymorphism was related to the risk of four digestive cancers (OR 1.315, 95% CI: 1.009-1.715), but such an association was not observed in stratified analysis and subgroup analysis except one study in Asian population (OR=3.104, 95% CI 1.923-5.011). Two studies indicated that the NAT2 polymorphism was associated with the risk of four cancers in Asian population (OR=1.701, 95% CI 1.019-2.838), and the association was also observed in smokers (OR=2.513, 95% CI 1.156-5.462). Moreover, smokers with the CYP1A1 Ile462Val polymorphism were at an increased cancer risk in Asian population (OR=1.585, 95% CI 1.029-2.442). None of the other gene-smoking interactions was observed in the above cancers.

Increasing studies investigated the gene-smoking interaction on the risk of cancer during these years. Two previously published studies indicated smokers with GSTM1 null genotype were at an increased oesophageal cancer risk 19, 21. Moreover, the significant association was found between CYP1A1 IIe462Val and liver cancer risk among the cigarette smoking subjects in a meta-analysis (OR = 1.40, 95% CI 1.06-1.85) 84. These results were similar to our findings. Zhang et al indicated the NAT2 polymorphisms were correlated to an increased liver cancer risk in smokers 11. Whereas our study only provided two studies to support this conclusion. The SULT1A1 Arg213His polymorphism was associated with an increased oesophageal cancer risk 85, but such an association was not founded in our subgroup analysis. We also found no interaction of smoking with other genetic polymorphisms on four digestive cancers. Several reasons account for the null results.

First, the association between gene polymorphism and cancer risk could be modified by various smoking habits, including the age of initiating smoking, duration of smoking, pack-years of smoking, the method of tobacco use and cigarette categories. One study showed that lifetime exposure to tobacco increased the risk of upper aero-digestive tract (UADT) cancers. Furthermore, chewing tobacco was more likely to increase the risk of UADT cancers (OR=7.61; 95% CI 4.65-12.45) compared to smoking 86. The categories of cigarette also play a role in cancer progression and affect the association of gene polymorphisms with cancer susceptibility 87. Remarkably, Liang et al reported on the significant interactions of smoking pack years with HEL308 genotypes (Pinteraction=0.026) and ADH1B genotypes (Pinteraction=0.0016) in the head and neck squamous cell carcinoma (HNSCC) risk, respectively 88. Most of the included studies only provided data to evaluate the smoking status and we could not verify the findings in our study. Moreover, the age of initiating smoking is rarely measured in published studies, but this factor could be related to genetic polymorphisms in subgroups. Second, many other genes could be relevant to the metabolism of harmful compounds in tobacco except for the included genes, and the gene-gene interaction also existed in cancer susceptibility 89, 90. It is probable that combinations of multiple gene polymorphisms are more significant as risk factors than a single gene polymorphism.

Interestingly, we found the GSTT1 null genotype could increase oesophageal cancer risk among non-smokers, but not among smokers. It was conflictive with the recognized conclusion on tobacco use increasing the cancer risk. However, this result also suggested not all the smokers with high-risk genetic variants were at an increased cancer risk. Because other benefical environmental factors, such as dietary habits, play an important role in cancer prevention 91. A previous study indicated that regular tea consumption decreased the OC (OR: 0.38, 95% CI: 0.17-0.87) and GC (OR: 0.30, 95% CI: 0.14-0.66) risk among those with GSTT1 null genotype 21. Ko et al also showed soy product consumption was associated with lower breast cancer risk in BRCA mutation carriers (HR: 0.39; 95% CI: 0.19-0.79) 92. It was resonalble to assume that a protective factor also interacted with the GSTT1 null genotype among smokers. Moreover, our finding was based on only 3 papers, and needed to be further verified by more studies.

Regarding the interaction between smoking and GSTM1 and CYP1A1 IIe462Val on digestive cancers risk, evidence regarding the molecular mechanism also supported the results of this meta-analysis. Tobacco smoke contains various carcinogens, for example, polycyclic aromatic hydrocarbons (PAHs) and tobacco specific nitrosamines (TSNA) 93. These carcinogens are first metabolically activated by phase I enzymes, e.g., cytochrome P4501A1 (CYP1A1), into their final forms and then combine with DNA, forming aromatic-DNA adducts that are considered as an early stage in carcinogenesis. Moreover, these activated forms are detoxified by phase II enzymes, especially glutathione S-transferases (GSTs)94. Thus, the susceptibility to cancer determined by genetic factors may depend on the metabolic balance between phase I and phase II enzymes8. Because the CYP and GST genetic polymorphisms regulate the metabolism of xenobiotics, they are thought to affect individual's sensitivity to environmental factors and susceptibility to cancer. Although this meta-analysis suggested that there was no significant interaction between smoking and other gene polymorphisms, several related molecular mechanisms remain biologically plausible. Except for the CYP and GST family genes, the carcinogens in tobacco smoke can be activated by SULT1A1 and NAT2 95, 96. DNA repair genes, e.g, hOGG1 and XRCC, are involved in the elimination of DNA adducts, which suggests that the DNA repair genes polymorphisms may be associated with different repair efficiencies of DNA damage 69. Moreover, the p53 is a tumour suppressor gene and plays a key role in regulating the cell cycle and maintaining genomic integrity 79. Thus, it may modify individual's susceptibility to various carcinogens.

Compared with a single study that investigated the role of some metabolic gene polymorphisms in cancer risk, we evaluated the interaction between ten gene polymorphisms and smoking for four digestive cancers, and this is the first such report to date. Therefore, we could provide more comprehensive information on the gene-smoking interaction in main digestive cancers. However, there are several limitations in this meta-analysis. First, there is strong heterogeneity in the risk estimates for most gene polymorphisms and stratified analyses. Second, the ORs were only adjusted for the cancer type and ethnicity. A more precise analysis should be performed based on the data adjusted for confounding factors including the age, sex, family history, environmental factors, cancer stage, and lifestyle. In addition, we were not able to evaluate the interaction of genes with genes or other environmental factors, which should be assessed in future studies.

In summary, our meta-analysis provides the evidence of two potential gene-smoking interactions, one is between smoking and GSTM1 on oesophageal cancer, and the other is between smoking and CYP1A1 Ile462Val on the four cancers in Asian populations. None of the other gene-smoking interactions was observed in the above cancer. Future studies need to be conducted to verify the conclusions.

Supplementary table S1.

Click here for additional data file.