Study screening results
In Fig. 1, 369 studies were selected through PubMed, EMBASE and Web of Science, with the help of the EndNote software, we excluded 118 repetitive articles. Reading the titles and abstracts of the articles, we found that 186 articles were clearly irrelevant. The full text of five studies was not found and we excluded 41 articles due to irrelevant content. Among these, 34 articles did not assess the acetylation status of NAT2, and 7 articles were reviews and meta-analysis. The Newcastle-Ottawa Scale (NOS) is the method we used to assess the quality of the remaining included studies. Table S1. shows the NOS scores for each study. Finally, there were 19 studies that met our criteria, including 4,130 lung cancer patients and 6,057 controls.
The major information we collected from these articles is shown in Table 1. The data collected included first author, year, country, ethnicity, source of controls, genotyping means, and phenotype distribution. The study by Sorensen M et al. was a nested case-control study, while the other studies were case-control studies. In addition, there were 6 hospital-based studies, 8 population-based studies, and 1 hospital and population-based study. In terms of patient backgrounds, 11 studies included European populations, and 8 included Asian populations. Of the genotyping methods, only one used TaqManSNP (TaqMan) [19].
Table 1
The data types extracted in this research
First author | Year | Country | Ethnicity | Source of controls | Genotyping method | Phenotype distribution Case Slow Rapid | Phenotype distribution Control Slow Rapid |
Mota[6] | 2015 | Portugal | European | HBa | PCR–RFLP | 153 36 | 168 61 |
Zhang[14] | 2014 | China | Asian | PBb | PCR–RFLP | 135 795 | 113 887 |
Mahasneh[7] | 2012 | Jordan | Asian | HB | PCR–RFLP | 22 27 | 61 38 |
Zupa[24] | 2009 | Italy | European | HB | PCR–RFLP | 43 32 | 50 71 |
Sobtic[15] | 2009 | India | Asian | HB | PCR–RFLP | 266 36 | 230 72 |
Lee[16] | 2009 | China | Asian | PB | PCR–RFLP | 22 95 | 36 83 |
Osawac[13] | 2007 | Japan | Asian | HB | PCR–RFLP | 53 60 | 50 71 |
Chen[17] | 2006 | China | Asian | PB | PCR–RFLP | 18 79 | 43 154 |
Borlak[9] | 2006 | UK | European | PB | PCR–RFLP | 39 28 | 152 91 |
Chiou[18] | 2005 | China | Asian | PB | PCR–RFLP | 27 135 | 64 144 |
Sorensenc[19] | 2005 | Denmark | European | Nestd | TaqMan PCR-RFLP | 243 12 | 239 25 |
Skuladottir[20] | 2005 | Denmark, Norway | European | PB | PCR–RFLP | 154 87 | 321 219 |
Belogubova[12] | 2005 | Russia | European | HB, PB | PCR–RFLP | 99 79 | 426 289 |
Wikman[21] | 2001 | German | European | HB | PCR–RFLP | 237 23 | 196 26 |
Hou[25] | 2000 | Norway | European | PB | PCR–RFLP | 169 112 | 237 138 |
Saarikoski[22] | 2000 | Finland | European | PB | PCR–RFLP | 102 93 | 152 140 |
Seow[10] | 1999 | China | Asian | HB | PCR–RFLP | 60 93 | 36 105 |
Nyberg[23] | 1998 | Sweden | European | PB | PCR–RFLP | 113 70 | 96 62 |
Cascorbi[26] | 1996 | Germany | European | PB | PCR–RFLP | 87 68 | 343 245 |
a. Hospital-based study. |
b. Population-based study. |
c. In the phenotype distribution of these three studies, slow represents the intermediate-slow acetylation status and fast represents the rapid acetylation status. |
d. Nested case-control study |
Meta-analysis
In the 16 studies used to analyze the slow versus rapid genotypes, we got the pooled ORs for slow versus rapid genotypes [6–7, 9–10, 12, 14, 16–18, 20–26]. As shown in Fig. 2A, the combined OR value of the 16 studies was 1.00 (95%CI: 0.84–1.19, I² = 63.3%, P < 0.001 for heterogeneity based on a random-effects analysis model). From the analysis of these 16 studies, we found no significant difference in lung cancer risk between rapid acetylators and slow acetylators. From this result, there was significant heterogeneity among the 16 studies. Therefore, in addition to taking smoking into account, we did a subgroup analysis of other factors, including ethnicity, source of controls, histological classification, and gender. The results are shown in Fig. 3. Metaninf is our method to investigate the influence of each study on the overall meta-analysis summary assessment [33]. The results are shown in Fig. 4. When we excluded the Chiou et al. [18]. study, the results differed the most from our original results. Based on the forest plot and metaninf analysis, we excluded the Chiou et al. [18], Seow et al. [10], Lee et al. [16] and Zupa et al. Articles [24]. After excluding these four studies, inter-study heterogeneity was significantly reduced (I² = 31.6%, p = 0.138). Excluding the four studies, the remaining 12 still did not show meaningful results (OR = 1.05, 95%CI: 0.94–1.17, as shown in the fixed-effects model). Figure 2B shows the results after excluding these four studies.
The Sobti et al. [15], Osawa et al. [13] and Sorensen et al. [19] studies divided the phenotypes of NAT2 into three types: the rapid type, intermediate type, and slow type, and Osawa's study classified the NAT2 intermediate phenotype and slow phenotype into one group. Therefore, we used these three studies to analyze whether there were significant differences in lung cancer risk among people with the rapid phenotype (homozygous rapid) and those with the other two acetylation statuses. As shown in Fig. 2A, the overall OR was 1.83 (95% CI: 1.22–2.73, I² = 39.6%, P = 0.191 for heterogeneity). The results showed a prominent decrease in the risk of lung cancer in those with the ultrarapid (homozygous rapid) phenotype, compared to those with the intermediate-slow phenotype.
Figure 5A shows the results obtained by including smoking as a subgroup factor. In the subgroup analysis of those who smoked versus those who never smoked, we did not find increased risks for either those who smoked (OR = 1.09, 95% CI = 0.64–1.85) versus those who never smoked (OR = 1.00, 95% CI = 0.65–1.54). From the results of the smoking and non-smoking groups, the heterogeneity between studies was significant; the I² of the smoking group was 74.8% and that of the non-smoking group was 70%. In addition, the smoking group heterogeneity P-value was 0.008, while the non-smoking group heterogeneity P-value was 0.005. When we excluded these two studies by Zhang et al. [14] in the smoking group and Chiou et al. [18] in the non-smoking group, the heterogeneity of the smoking and non-smoking studies was significantly reduced (smoking group: I²<0.1%, P = 0.824; non-smoking group: I² = 25.3%, P = 0.253). After elimination, the smoking group had an OR = 0.84, 95%CI: 0.62–1.15, and the non-smoking group had an OR = 1.20, 95%CI: 0.88–1.62. Figure 5B shows the results after excluding these two studies. The results showed that in both the smoking and non-smoking groups, there was no significant augment in the risk of lung cancer among slow genotypes compared to rapid genotypes. We show the subgroup analysis results of other factors in Fig. 3. In the stratified analysis, we found no statistically significant correlation. The major results of the heterogeneity test are also shown in Table 2, indicating that the heterogeneity of the European population (I2 = 26.1%, P = 0.203) and female (I2 = 50.9%, P = 0.086) participants were significantly reduced.
Table 2
Results of each subgroup analysis
Subgroups | Na | OR | 95%CI | I2 | Pb |
Source of controls | | | | | |
HB | 5 | 1.32 | (0.91, 1.91) | 65.7% | 0.020 |
PB | 11 | 0.91 | (0.76, 1.08) | 53.7% | 0.017 |
Ethnicity | | | | | |
European | 10 | 1.06 | (0.93, 1.22) | 26.1% | 0.203 |
Asian | 6 | 0.82 | (0.50, 1.33) | 82.3% | < 0.001 |
Gender | | | | | |
male | 4 | 0.88 | (0.51, 1.50) | 65.6% | 0.033 |
female | 5 | 1.04 | (0.65, 1.66) | 50.9% | 0.086 |
Histological type | | | | | |
Adenocarcinoma | 5 | 1.01 | (0.68, 1.49) | 62.3% | 0.031 |
Squamous carcinoma | 4 | 0.79 | (0.53, 1.19) | 63.4% | 0.042 |
Small-cell carcinoma | 2 | 0.95 | (0.57, 1.60) | < 0.1% | 0.366 |
Large-cell carcinoma | 2 | 2.27 | (0.96, 5.37) | < 0.1% | 0.590 |
Totalc | 16 | 1.00 | (0.84, 1.19) | 63.3% | < 0.001 |
a. The number of studies. |
b. The p-value for heterogeneity. |
c. The 16 studies used to analyze the slow versus rapid genotypes. |