Smoking status and subsequent gastric cancer risk in men compared with women: a meta-analysis of prospective observational studies

This systematic review was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) Statement issued in 2009 [13]. Relevant articles were systematically searched in MEDLINE, EMBASE, and the Cochrane CENTRAL electronic databases from database inception to December 2018. We included studies that investigated humans without language restrictions and regardless of publication status (published, in the press, or in progress). The studies reporting associations between smoking status and gastric cancer risk were searched using strategies of a combined text and medical subjects headings (MeSH): (“smoke” OR “smoking” OR “nicotine” OR “tobacco” OR “lifestyle” OR “lifestyles” OR “cigarette”) AND (“gastric” OR “stomach” OR “cardia”) AND (“cancer” OR “tumor” OR “neoplasm”) AND (“nested case control” OR “cohort” OR “prospective”). Furthermore, we also manually checked the reference lists of identified reports for other potentially relevant studies. If the same population was reported more than once, the most comprehensive and recently published article was used. The study topic, study design, exposure, population, and reported outcomes were used to identify relevant studies.

Two authors independently performed a literature search and study selection, and disagreements between two authors were settled by a discussion in a group until a consensus was reached. A study was deemed eligible if it met the following inclusion criteria: (1) the study design was a prospective observational study; (2) the study evaluated the association of smoking status with gastric cancer risk; and (3) the associations between smoking status and gastric cancer risk in men and women were both reported.

Two authors independently collected and extracted data from the included studies, and disagreements were resolved by a group discussion. The data collected from the included studies contained the following items: first author, publication year, country, sex, sample size of men and women, mean age for men and women, number of participants who had never smoked (non-smokers) for men and women, number of former smokers in men and women, number of current smokers in men and women, follow-up duration, reported outcomes, and adjusted factors.

We assessed the methodological quality of the included studies using the Newcastle-Ottawa Scale (NOS) [14], which has been partially validated for evaluating the quality of observational studies included in meta-analyses. The NOS is based on selection (4 items), comparability (1 item), and outcome (3 items), and provides a “star system” range of 0–9 to evaluate study quality. Two authors independently performed quality assessments and disagreements were settled by a group discussion.

The associations between smoking status and gastric cancer risk in men and women were determined based on the relative risk (RR), hazard ratio (HR), or odds ratio (OR), and the 95% confidence intervals (CIs) in each individual study. HR is considered equivalent to RR in prospective observational studies, and OR could also be assumed to be equivalent to the RR due to the low incidence of gastric cancer. We calculated the ratio of RRs (RRR) for current or former smokers versus non-smokers and the risk of gastric cancer based on sex-specific RRs in individual studies [15]. We used random-effects models to calculate the summary RRR and compared the sex differences in gastric cancer risk in current smokers, former smokers, or non-smokers [16, 17].

Heterogeneity among studies was shown by the I² and Q statistics, and P values < 0.10 mean significant heterogeneity [18, 19]. A sensitivity analysis was performed by systematically excluding each study individually to evaluate its influence on the meta-analysis [20]. The potential sources of heterogeneity in estimates of the impact of current and former smokers based on follow-up duration were explored by using univariate meta-regression [21]. Subgroup analyses for the sex differences in the association between smoking status and gastric cancer risk were based on publication year, country, follow-up duration, reported outcomes, whether or not the studies adjusted for BMI or alcohol consumption, and study quality. Publication bias was explored visually using funnel plots and statistically using Egger’s and Begg’s tests [22, 23]. All P values were two-sided with significance defined as P < 0.05. Statistical analyses were conducted using STATA software (version 10.0; Stata Corporation, College Station, TX, USA).

A total of 1517 records from the initial search were identified, including 691 from MEDLINE, 757 from EMBASE, and 69 from the Cochrane CENTRAL. After discarding 1423 irrelevant or duplicate studies, 94 potential studies were selected for further reading. After detailed evaluating, 10 prospective observational studies were selected into the quantitative analysis [24‐33]. The manual search of the reference lists of these studies did not yield any new eligible studies. The systematic review selection process is shown in Fig. 1, and the general characteristics of the included studies are displayed in Table 1.

Ten studies with a total of 3,381,345 participants were included in our analysis. Among the studies, nine were prospective cohort studies [24‐27, 29‐33] and one was a nested case-control study [28]. The duration of follow-up for participants was 5.0–28.0 years, while 9753-1,212,906 individuals were included in each study. Three studies were conducted in Japan [24, 25, 31], one in Korea [30], two in Norway [26, 27], one in the UK [28], two in the US [30, 33], and one in 10 European countries [32]. The main study outcome in 6 studies was gastric cancer mortality, and the remaining 4 studies reported gastric cancer incidence. NOS scores were used to evaluate study quality [14], and a score ≥ 7 was regarded as high quality. Overall, three studies had scores of 8, four studies had scores of 7, and the remaining three studies had scores of 6.

All included studies reported sex differences in the association between gastric cancer risk and current smokers compared with non-smokers. We noted current smokers were associated with higher risk of gastric cancer when compared with non-smokers in men (RR: 1.63; 95% CI: 1.44–1.85; P < 0.001; Fig. 2) and women (RR: 1.30; 95% CI: 1.06–1.60; P = 0.010; Fig. 2). Further, the increased risk of gastric cancer in current smokers compared to non-smokers was higher in men than in women (RRR: 1.30; 95% CI: 1.05–1.63; P = 0.019; Fig. 3), with significant heterogeneity (I² = 52.6%; P = 0.025). The result of the sensitivity analysis indicated that the sex differences in the association between current smokers and gastric cancer were affected by the exclusion of multiple studies due to the small numbers of cohorts included (Table 2). The results of the meta-regression analysis showed that follow-up duration was not a significant factor contributing to the sex differences of the association between current smokers and gastric cancer (Additional file 1). We used subgroup analyses to minimize heterogeneity among the included studies and evaluate the sex differences in subpopulations (Table 3). The summary RRR (male to female) for current smokers indicated an increased risk of gastric cancer in men when the study was conducted in Asia (RRR: 1.50; 95% CI: 1.17–1.91; P = 0.001), regardless of follow-up duration (follow-up duration ≥10.0 years [RRR: 1.33; 95% CI: 1.02–1.74; P = 0.037]; follow-up duration < 10.0 years [RRR: 1.46; 95% CI: 1.11–1.91; P = 0.006]), when the study reported gastric cancer mortality (RRR: 1.53; 95% CI: 1.24–1.89; P < 0.001), when the study did not adjust for BMI (RRR: 1.47; 95% CI: 1.24–1.74; P < 0.001), when the study did not adjust for alcohol consumption (RRR: 1.53; 95% CI: 1.20–1.94; P = 0.001), and when the study had a NOS score of 7 or 8 (RRR: 1.42; 95% CI: 1.11–1.81; P = 0.005).

A total of 9 studies reported sex differences in the relation between gastric cancer risk in former smokers compared to non-smokers. The summary result indicated former smokers were associated with an increased risk of gastric cancer in men (RR: 1.42; 95% CI: 1.31–1.54; P < 0.001; Fig. 4), while this association was not associated with statistically significant in women (RR: 1.19; 95% CI: 0.96–1.47; P = 0.112; Fig. 4). There was no significant difference for gastric cancer risk between former smokers and non-smokers in men compared with women (RRR: 1.20; 95% CI: 0.92–1.55; P = 0.178; Fig. 5), and potential significant heterogeneity was observed among the included studies (I² = 41.8%; P = 0.089). Following the result of the sensitivity analysis, we excluded the study by Lindblad et al. [28], which used a nested case control design. After this exclusion, we could conclude that male former smokers had a significantly increased risk of gastric cancer over non-smokers compared to female former smokers (RRR: 1.31; 95% CI: 1.10–1.57; P = 0.002; Table 2). Meta-regression analysis indicated follow-up duration did not contribute a significant role with the sex difference of the relation between former smokers and gastric cancer (Additional file 1). Subgroup analyses indicated a higher risk of gastric cancer in male verses female former smokers when the study was conducted in Asia (RRR: 1.35; 95% CI: 1.05–1.74; P = 0.019; Table 4), follow-up duration < 10.0 years (RRR: 1.35; 95% CI: 1.02–1.80; P = 0.038), when the study reported gastric cancer mortality (RRR: 1.25; 95% CI: 1.03–1.51; P = 0.022; Table 4), when the study did not adjust for BMI (RRR: 1.26; 95% CI: 1.04–1.53; P = 0.019; Table 4), when the study did not adjust for alcohol consumption (RRR: 1.26; 95% CI: 1.04–1.53; P = 0.020; Table 4), and when the study had high study quality (RRR: 1.32; 95% CI: 1.09–1.59; P = 0.004; Table 4). Furthermore, male former smokers were associated with a lower risk of gastric cancer if the study had lower study quality (RRR: 0.34; 95% CI: 0.12–0.93; P = 0.036).

Reviewing the funnel plots could not rule out the potential publication bias contributing to the sex differences in gastric cancer risk. The Egger’s and Begg’s test results showed no evidence of publication bias for sex differences in the association between current smokers and gastric cancer risk (Fig. 6). Moreover, there was no significant publication bias for former smokers and gastric cancer risk (Fig. 7).