Introduction

Tobacco smoking and alcohol consumption were considered as lifestyle-related risk factors correlated with increased risk of upper aerodigestive tract cancer (UADTC) with synergistic effect [20, 22]. The carcinogenic effects of tobacco and alcohol were identified to be multiplicative on the relative risk magnitude in a meta-regression [41]. Tobacco appeared to possess a much stronger impact on the larynx than on the other aerodigestive sites [22]. The heterogeneity of the pooled estimates was present among researches with or without adjustment for alcohol consumption in a meta-analysis based on information reported in the International Agency for Research on Cancer (IARC) Monograph on ‘‘Tobacco Smoke and Involuntary Smoking’’ [15]. Lung cancer (RR = 8.96; 95% CI 6.73–12.11) and laryngeal cancer (RR = 6.98; 95% CI 3.14–15.52) showed the highest relative risks for current smokers. Evidence from either observational or experimental researches had presented a positive correlation between cigarette smoking and risk of incidence or mortality of laryngeal cancer [19].

The association between cigarette smoking and laryngeal cancer risk had been reviewed in several publications [38]. The pooled estimate of 120 studies for the association between current smoking and risk of UADTC was 3.47 (95% CI 3.06–3.92) in a meta-analysis, with the strength of the association being significantly stronger for laryngeal cancer (RR 9.07; 95% CI 6.33–13.0) versus other subtypes of UADTC [30]. A strong dose–response association between the number of cigarettes smoked per day and the risk of laryngeal cancer was observed both before and after adjustment for publication bias [30]. A pooled analysis of case–control studies indicated that greater number of cigarettes smoked per day for a shorter duration was less deleterious than fewer cigarettes/day for a longer duration and the greater risk of laryngeal cancer derived from different numbers of cigarettes smoked per day effects and not pack-years [27].

However, few studies examined the association between cigarette smoking status, pack-years, frequency, duration and cessation of cigarette smoking and laryngeal cancer risk through a meta-analysis. In this study, evidence from published observational researches regarding the association between cigarette smoking without alcohol consumption and laryngeal cancer risk were searched and summarized quantitatively with a meta-analytic approach. The study was carried out and written following the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines.

Materials and methods

Study identification, eligibility and exclusion criteria

All published epidemiologically observational studies which surveyed the correlation between cigarette smoking and risk of laryngeal cancer were included in the meta-analysis if they satisfied all of the following criteria: (1) possessed original data expressed as pooled estimates of relative risks (RRs) and their 95% confidence intervals(CI) from case–control studies, nested case–control studies or cohort studies; (2) the primary outcome of cases was clearly diagnosed as the cancer of the larynx; (3) had a quantitative estimate of the association between cigarette smoking status, pack-years, frequency, duration and cessation of cigarette smoking and laryngeal cancer risk, expressed as hazard ratios(HRs for cohort studies) and odds ratios (ORs for case–control studies) for each category of cigarette smokers versus never smokers; (4) risk estimates were necessarily adjusted for selected confounding factors or covariates; (5)were published in English or Chinese language; (6) subjects and controls did not use alcohol, and (7) were published up to March 2016. If a study arised in various articles, data with the largest sample size published in the most recent years were included for the meta-analysis.

The pooled analysis, reviews, meta-analysis, duplicated, unpublished or non-original researches, cross-sectional and articles in cell lines or animals were excluded. In addition, studies were excluded from this meta-analysis if they met any one of the following criteria: (1) laryngeal cancer cases were concurrently suffered from viral infection, occupational factors exposure, obesity, and other diseases such as diabetes, concurrent or secondary cancers; (2) other forms of tobacco-related consumption, including tobacco chewing, chewing of betel leaf with tobacco, smokeless tobacco, water pipes, cigar, cigarillos or bidi smoking, pipe smoking, marijuana smoking, electronic cigarette, involuntary smoking exposure at home or at work; (3) cases with alcohol consumption, betel quid or areca nut chewing, and a family of malignancy; (4) cases were previously under radiotherapy, chemotherapy and surgical treatment; (5) not reporting a quantitative estimate of the association expressed as pooled estimates of ORs or RRs and their 95% CI.

Search strategy

Such databases as Up to Date, ACP journal club, Medline, PubMed Clinical Queries, Sumsearch, Ovid, Tripdatabase, Cochrane Library (CDSR, CCTR, DARE), Embase, Chinese Biomedical Database(CBM), China National Knowledge Infrastructure(CNKI), CqVip, and Wanfang were completely searched from their establishment to March 2016 for gathering relevant medical literature. Data retrievals were limited to the English and Chinese language and done in parallel by Chen Chen and Jing Jing Zuo. Free term searching was adopted given that observational studies were not indexed as subject terms in the search engines. The key words and free term used were as follows: (“cigarette smoking” OR “tobacco smoking”) AND (“laryngeal cancer” OR “head and neck cancer”) AND (etiology OR “risk factor”) AND (“case–control study” OR “cohort study”). The articles reporting data on the association between cigarette smoking and risk of laryngeal cancer incidence were selected, with no other and previous diagnosis of cancer at any site of body. Figure 1 shows the flowchart for screening of articles. The reference lists of all papers of interest were examined to get other related publications and the corresponding authors were contacted to confirm data extraction when data were incomplete or uncertain data extraction and methodological quality assessment.

Fig. 1
figure 1

The flowchart for selection of articles

Full size image

Baseline information of available data were drawn and compared by two reviewers separately using the same criteria. The disagreements about results were resolved through consensus. Information on country, design, sample size, and sex of studies, the variables that the study results were controlled, the RRs and their corresponding 95% CIs for cigarette smoking status and each category of the pack-years, frequency, duration and cessation of cigarette smoking were extracted. The maximal results adjusted were adopted from the studies which reported both crude and adjusted RRs and 95% CIs, while the papers with unadjusted estimates only were ruled out. The more updated article was contained when two or more papers offered results from the same research. The countries or districts of included studies were divided into Europe–America and non-Europe–America, while amounts of sample sizes were categorized as big sample size (≥300) and small sample size (<300). Quality evaluation of included studies was conducted following the Newcastle–Ottawa Scale (NOS) criteria which possessed four items for selection, two items for comparability, and three items for evaluation of outcomes or exposures. A maximum of nine points was designated to each research, while scores of 7–9 and 0–6 were considered as high quality and non-high quality separately.

Statistical analysis

The adjusted relative risks (RRs) and their corresponding 95% CI from each eligible research were combined for the meta-analysis, weighted by the inverse of their variance, and p < 0.05 was considered statistically significant. The pooled RRs and 95% CIs between laryngeal cancer and cigarette smoking were estimated by meta-analysis and offered with forest plots for each category of cigarette smokers versus never smokers. Meanwhile, the association between laryngeal cancer risk and the pack-years, frequency, duration and cessation of cigarette smoking was carried out separately. Statistical heterogeneity among studies was assessed with I 2 statistic and the results with p < 0.10 were identified as heterogeneous [21]. The results of I 2 statistic with values of 25, 50, and 75% represented low, moderate, and high degrees of heterogeneity, respectively. Heterogeneity was regarded statistically significant when I 2 > 50% and p < 0.1, then sensitivity analysis was conducted by excluding one study once each time and investigating the impact of each individual research on the overall relative risk. Subgroup analysis was implemented to find out sources of research heterogeneity and the effect of potential confounding factors such as country, design, sample size, and sex of studies.

Heterogeneity was not recognized statistically significant when I 2 < 50% and p > 0.1 and fixed effects model was adopted. Instead, random effects model that took variation both between and within researches into account was then chosen to analyze the data. Publication bias was assessed through examination of Egger’s test and Begg’s test [8] (significant at p < 0.1). Trim and fill algorithm was conducted to check and correct for the asymmetry of funnel plot from publication bias possibly [35]. All statistical analyses were conducted with the Stata 11 (Stata Corporation, College Station, TX, USA) statistical software and their tests are two-sided.

Results

A total of four cohort studies and 26 case–control studies, with information about 14,292 cases of laryngeal cancers and 45,579 individuals as control, were included in the meta-analysis [1–7, 9–14, 17, 18, 23–26, 28, 31–34, 36, 37, 39, 40, 42, 43]. The summary features of included studies with risk estimates and corresponding 95% CI for cigarette smoking and laryngeal cancer are showed in Table 1. The majority of the researches were from Europe (n = 12), America (n = 6), Asia (n = 11), and Africa (n = 1). Table 2 gave the risk estimates and their corresponding 95% CIs for each category of the pack-years, frequency, duration and cessation of cigarette smoking and laryngeal cancer.

Table 1 Main characteristics of the studies on laryngeal cancers and cigarette smoking included in the meta-analyses
Full size table
Table 2 Meta-analyses and subgroup analyses of characteristics of cigarette smoking and risk of laryngeal cancer (cigarette smokers versus never smokers)
Full size table

Quality assessment

Quality evaluation of all included studies was carried out using the Newcastle–Ottawa Quality Assessment Scale for case–control and cohort studies (Table 1). Among case–control studies, 16 studies (accounted for 61.5%) were considered as high quality and only Pacella-Norman et al. [32] scored the lowest, while all cohort studies scored highest independently. Most studies were of good quality with no evidence of selection bias, good outcome assessment of cohort studies, and with good comparability of the exposed and unexposed groups of case–control studies.

Cigarette smoking status and the risk of laryngeal cancer

Twenty-four articles that presented the correlation between cigarette smoking and risk of laryngeal cancer were identified and their pooled estimate was 6.62 (95% CI 5.26–8.35) through a meta-analytic approach, with significant heterogeneity between the studies (I 2 = 83.5%, p = 0.000). Sensitivity analysis via excluding every study was conducted to find out six related data of studies [8, 9, 12, 16, 20, 26] as the source of heterogeneity between study estimates (Fig. 7). The subsequent analyses were limited to studies (n = 18) involving 3279 laryngeal cancers events and 11,252 participants. The pooled estimate of the eighteen studies for the correlation between current smoking and risk of laryngeal cancer was 7.01 (95% CI 5.56–8.85), with moderate heterogeneity across the studies (I 2 = 56.7%, p = 0.002; Table 2). The RRs were 5.04 (95% CI 3.09–8.22) for cohort studies (p = 0.121), 7.59 (95% CI 5.86–9.82) for case–control studies (p = 0.005; Fig. 2). Sensitivity analysis by omitting every research revealed similar outcomes, suggesting the robustness of the results.

Fig. 2
figure 2

Cigarette smoking status and the risk of laryngeal cancer in a meta-analysis. Black squares represent point estimates and horizontal lines indicate 95% CIs for the observed effect in each research. A white diamond represents a pooled estimate and 95% CI for meta-analysis. ID1 cohort study, ID2 case–control study, RR relative risk, CI confidence interval

Full size image

A great amount of the heterogeneity was interpreted by differences between the two types of sample size using a subgroup analysis method, with the strength of the association being notably stronger for small sample size group(RR 7.19; 95% CI 5.31–9.74) versus large sample size group (RR 6.46; 95% CI 5.05–8.27). The research heterogeneity was much lower in studies with large sample size (I 2 = 2.5%, p = 0.359), compared with studies with small sample size (I 2 = 62.2%, p = 0.001; Table 2). The different heterogeneity between the two types of sample size once again confirmed that the studies with small sample size, compared with those with large sample size, had lower quality statistically and were prone to show a higher effect estimate leading to bias.

Geographical discrepancies of the research populations were another source of heterogeneity between study estimates, with the extent of the association being obviously greater among populations from Europe and America group compared with non-Europe and America group: the RRs were 7.89 (95% CI 5.77–10.77) and 6.64 (95% CI 4.83–9.12) separately. The study heterogeneity was much lower in studies from Europe and America group (I 2 = 24.1%, p = 0.237) compared with that in studies from non-Europe and America group (I 2 = 87.2%, p = 0.001), suggesting inferior quality of the latter studies relatively.

Information on sex of studies was extracted and moderate heterogeneity was shown in one sex group using a subgroup analysis method (I 2 = 25.3%, p = 0.227). Comparing with the size of heterogeneity in geographical discrepancies or sample size, the differences in sexes were not the primary source of heterogeneity between study estimates.

Pack-years of cigarette smoking and risk of laryngeal cancer

Twenty-seven studies reported the association between pack-years of cigarette smoking and risk of laryngeal cancer and their pooled risk estimate was 4.31 (95% CI 3.22–5.77), with prominent heterogeneity between the studies (I 2 = 95.2%, p = 0.000). Sensitivity analysis through excluding every study was adopted to identify five related data of studies [18, 34] as the source of heterogeneity between study estimates.

A total of 22 articles had presented four proprietary forms for pack-years of smoking: never smokers, <20 pack-years, 20–39 pack-years, and ≥40 pack-years, including 2367 laryngeal cancers events and 7613 participants. The summary estimate of the 22 identified studies for the association between pack-years of cigarette smoking and risk of laryngeal cancer was 4.61 (95% CI 3.44–6.18), with moderate heterogeneity between the studies (I 2 = 62.7%, p = 0.000; Fig. 3) and no significant heterogeneity between the four subtypes of studies categorized by pack-years (Table 2). The RRs were 3.03 (95% CI 1.88–4.89) for cohort studies (p = 0.326), 4.97 (95% CI 3.58–6.92) for case–control studies (p < 0.001).

Fig. 3
figure 3

Pack-years of cigarette smoking and risk of laryngeal cancer in a meta-analysis (for definitions of the squares, lines, and diamond; see Fig. 1). ID1 0–19 pack-years, ID2 20–39 pack-years, ID3 ≥40 pack-years, RR relative risk, CI confidence interval

Full size image

The risk of laryngeal cancer exacerbated gradually as the number of pack-years of cigarette smoking increased. Individuals who smoked with 40 or more pack-years had ninefold the risk of laryngeal cancer versus never smokers (RR 9.14; 95% CI 6.24–13.39). Sensitivity analysis by deleting every research showed similar results, indicating the robustness of the outcomes. The moderate heterogeneity was explained by differences between the designs of study, with the strongpoint of the association being stronger for case–control group (RR 4.97; 95% CI 3.58–6.92) versus cohort group (RR 3.03; 95% CI 1.88–4.89). The study heterogeneity was much lower in cohort group (I 2 = 10.7%, p = 0.326; Table 2), compared with that in case–control group (I 2 = 65.2%, p = 0.000). The outcomes of heterogeneity test demonstrate that the results of cohort studies, which had adjusted for major confounding factors, were more reliable than that of retrospective case–control studies which were more subject to bias.

Frequency of cigarette smoking and risk of laryngeal cancer

Thirty studies had investigated four specific types for daily cigarette consumption: never smokers, <20 cigarettes per day, 20–29 cigarettes per day, and ≥30 cigarettes per day. The pooled estimate of these articles that presented the association between frequency of cigarette smoking and risk of laryngeal cancer was 6.34 (95% CI 4.23–9.48), with noteworthy heterogeneity between the studies (I 2 = 86.5%, p = 0.000). Sensitivity analysis by eliminating every research identified nine related data of [9, 12, 16, 20, 26] studies as the source of heterogeneity between study estimates.

The resulting analyses were confined to twenty-one studies involving 2814 laryngeal cancer cases and 9636 controls. The pooled estimate of these studies for the association between frequency of cigarette smoking and risk of laryngeal cancer was 5.82 (95% CI 4.47–7.58), with medium heterogeneity between the studies (I 2 = 49.9%, p = 0.005; Fig. 4). The RRs were 3.65 (95% CI 2.40–5.54) for cohort studies (p = 0.523), 6.53 (95% CI 4.85–8.78) for case–control studies (p = 0.008). A substantial dose–response correlation between the number of cigarettes smoked per day and risk of laryngeal cancer was surveyed, such that the risk remained high for smoking <30 cigarettes a day and individuals who smoked 30 or more cigarettes a day had seven times the risk of laryngeal cancer versus never smokers (RR 7.02; 95% CI 4.47–11.02; Table 2). Sensitivity analysis by omitting every research revealed similar results, showing the robustness of the outcomes.

Fig. 4
figure 4

Frequency of cigarette smoking and risk of laryngeal cancer in a meta-analysis (for definitions of the squares, lines, and diamond; see Fig. 1). ID1 0–19 cigarettes per day, ID2 20–29 cigarettes per day, ID3 ≥30 cigarettes per day, RR relative risk, CI confidence interval

Full size image

A large amount of the heterogeneity was interpreted by differences between continents adopting a subgroup analysis method, with the strength of the association being notably stronger for non-Europe and America group (RR 6.48; 95% CI 4.72–8.89) versus Europe and America group (RR 5.72; 95% CI 4.04–8.10). The study heterogeneity was much lower in non-Europe and America group (I 2 = 0.0%, p = 0.728), compared with that in Europe and America group (I 2 = 59.2%, p = 0.001; Table 2). Discrepancies between the design of studies were the other source of heterogeneity between study estimates, with the dimension of the association being stronger among populations from case–control group compared with cohort group: the RRs were 6.53 (95% CI 4.85–8.78) and 3.65 (95% CI 2.40–5.54) severally. The study heterogeneity was relatively lower in cohort group (I 2 = 0.0%, p = 0.523), compared with that in case–control group (I 2 = 51.0%, p = 0.008).

Duration of cigarette smoking and risk of laryngeal cancer

Twenty-nine researches presented the association between duration of cigarette smoking and risk of laryngeal cancer and their summary estimate by meta-analysis was 4.52 (95% CI 3.21–6.37), with significant heterogeneity between the studies (I 2 = 84%, p = 0.000). Sensitivity analysis by omitting every research was used to find out ten related data of studies [8, 9, 12, 16] as the source of heterogeneity between study estimates. The final analyses were restrained to studies (n = 19) involving including 2699 laryngeal cancer cases and 9669 controls. The pooled estimate of the 19 reports for the association between duration of cigarette smoking and risk of laryngeal cancer was 4.27 (95% CI 3.62–5.05; Table 2), with slight heterogeneity between the studies (I 2 = 26.9%, p = 0.136, fixed effects model used; Fig. 5). The RRs were 2.87 (95% CI 2.09–3.96) for cohort studies (p = 0.560), 4.95 (95% CI 4.07–6.01) for case–control studies (p = 0.398).

Fig. 5
figure 5

Duration of cigarette smoking and risk of laryngeal cancer in a meta-analysis (for definitions of the squares, lines, and diamond; see Fig. 1). Overall relative risk calculated with fixed effects model. ID1 0–19 years, ID2 20–29 years, ID3 30–39 years, ID4 ≥40 years, RR relative risk, CI confidence interval

Full size image

A strong time–response correlation between duration of cigarette smoking and the risk of laryngeal cancer was indicated, such that the risk kept raised within 40 years and individuals who smoked 40 or more years had fivefold the risk of laryngeal cancer versus never smokers (RR 5.76; 95% CI 3.69–8.99; Table 2). Sensitivity analysis by excluding every research indicated similar outcomes, showing the robustness of the outcomes. Differences between the published years of studies might be the source of heterogeneity between study estimates through subgroup analysis, with the extent of the correlation being greater among studies published before 2006 compared with studies published after 2006: the RRs were 4.50 (95% CI 3.71–5.45) and 3.67 (95% CI 2.64–5.12) separately. The study heterogeneity was obviously lower in studies published after 2006 (I 2 = 0.0%, p = 0.524), compared with that in studies published before 2006 (I 2 = 42.5%, p = 0.066).

Smoking cessation and risk of laryngeal cancer in former smokers

Thirteen studies that presented the association between years of quitting smoking and risk of laryngeal cancer were defined and their summary estimate was 1.09 (95% CI 0.53–2.24), with significant heterogeneity between the studies (I 2 = 93.8%, p = 0.000). Sensitivity analysis via excluding every research was conducted to identify six related data of [9, 16, 18, 36] studies as the source of heterogeneity between study estimates. The subsequent analyses were limited to case–control studies (n = 7) involving 924 laryngeal cancer events and 2902 participants. The pooled estimate of the seven defined studies for the association between years of smoking cessation and risk of laryngeal cancer was 2.37 (95% CI 1.60–3.50; Table 2), with slight heterogeneity between the studies (I 2 = 22.9%, P = 0.255, fixed effects model adopted; Fig. 6).

Fig. 6
figure 6

Smoking cessation and risk of laryngeal cancer in former smokers in a meta-analysis (for definitions of the squares, lines, and diamond; see Fig. 1). Overall relative risk calculated with fixed effects model. ID1 0–15 years, ID2 ≥16 years, RR relative risk, CI confidence interval

Full size image

The risk of laryngeal cancer kept elevated within the first 15 years of quitting smoking (RR 3.62, 95% CI 1.88–7.00) but declined in the 16 years and more after smoking cessation (RR 1.88, 95% CI 1.16–3.05, I 2 = 18.8%, p = 0.296, fixed effects model used). Sensitivity analysis by excluding every study revealed similar outcomes, suggesting the robustness of the outcomes. Subgroup analysis was not conducted to investigate source of study heterogeneity and the impact of latent confounding factors since this meta-analysis included under ten studies.

Influence of publication bias on the strength of the association

Publication bias was evaluated by visual inspection of a funnel plot, Egger’s test and Begg’s test (significant at p < 0.1; Figs. 7, 8). Evidence of publication bias was not detected for the association between current cigarette smoking (versus never smoking) and risk of laryngeal cancer (p = 0.225 with Begg’s test, p = 0.317 with Egger’s test; Table 2). Moreover, publication bias was not observed for studies presenting on the association between pack-years of cigarette smoking and risk of laryngeal cancer (p = 0.310 with Begg’s test, p = 0.110 with Egger’s test), the association between the number of cigarettes smoked per day and risk of laryngeal cancer(p = 0.566 with Begg’s test, p = 0.327 with Egger’s test), and the correlation between duration of cigarette smoking and risk of laryngeal cancer (p = 0.726 with Begg’s test, p = 0.844 with Egger’s test), respectively. Egger’s test and Begg’s test were not adopted to investigate publication bias for studies presenting on the association between years of quitting smoking and the risk of laryngeal cancer since this part of meta-analysis consisted of under ten studies.

Fig. 7
figure 7

Publication bias on cigarette smoking status and the risk of laryngeal cancer was assessed through examination of a Begg’s funnel plot

Full size image
Fig. 8
figure 8

Publication bias on cigarette smoking status and the risk of laryngeal cancer was assessed through examination of a Egger’s test

Full size image

Discussion

The present meta-analysis including all published eligible data identified a significant increase in the risk of laryngeal cancer in cigarette smokers compared with never smokers. The results of this meta-analysis including 14,292 laryngeal cancer events from 30 studies demonstrated strong association referring to dose–response and time–response between cigarette smoking and risk of laryngeal cancer for both men and women. The risks of laryngeal cancer were especially evident for individuals who smoked with 40 or more pack-years, subjects who smoked 30 or more cigarettes a day, and persons who smoked 40 or more years. The risks remained elevated for 15 years after smoking cessation but dropped afterwards. The vital public health implication was that the chance of developing laryngeal cancer might be distinctly decreased by quitting smoking, particularly among former cigarette smokers who had stopped smoking for 15 or more years.

The results of the meta-analysis pointing out the strong association between cigarette smoking and risk of laryngeal cancer supported the argument that cigarette smoking was more strongly associated with laryngeal cancer and the association was not influenced by the differences in sexes of subjects and controls [16, 30]. Comparing with the size of heterogeneity in geographical discrepancies or sample size, the differences in sexes were not the primary source of heterogeneity between study estimates. However, contrary to previous viewpoint that the great laryngeal cancer risk with smoking derived from the differential effects of cigarettes/day and not pack-years [27], this meta-analysis suggested that pack-years, frequency and duration of cigarette smoking were associated with a 4.6-fold, 5.8-fold, and 4.2-fold increased risk of laryngeal cancer, respectively. The pooled estimate of the 18 studies for the association between current smoking and risk of laryngeal cancer was 7.01 (95% CI 5.56–8.85), which was less than the result (RR 9.07; 95% CI 6.33–13.0) had reported [30]. The risks remained high for a 15 years after smoking cessation but dropped afterwards, compared with the previous view that the risk of upper aerodigestive tract cancer remained elevated in the first decade after smoking cessation but declined thereafter. The extent of the correlation was obviously greater among populations from Europe and America group compared with non-Europe and America group: the RRs were 7.89 (95% CI 5.77–10.77) and 6.64 (95% CI 4.83–9.12) separately, compared with the previous result that the difference corrected for the presence of publication bias between regions was reduced and became nonsignificant: the RR was 2.39 (95% CI 1.98–2.89) in Asian countries and 3.10 (95% CI 2.50–3.85) in non-Asian countries.

The main advantages of the present meta-analysis were based on a large number of studies and the first quantitative evaluation with respect to the association between laryngeal cancer risk and pack-years, frequency, duration and cessation of cigarette smoking. Cigarette smoking was regarded as the most vital risk factor for laryngeal cancer distinctly on the basis of the complete review of observational studies.

Sensitivity analysis was conducted by subgroup analysis and by excluding one study every time, investigating the effect of each study on the overall relative risk. The results of pooled relative risks and their 95% confidence intervals were changed after the first sensitivity analysis, while orientations of results were consistent. Most studies were of good quality with no evidence of selection bias, good outcome assessment of cohort studies, and with good comparability of the exposed and unexposed groups of case–control studies. Evidence of publication bias was not detected for the association between current cigarette smoking (versus never smoking) and risk of laryngeal cancer. Egger’s test and Begg’s test were not adopted to investigate publication bias for seven studies presenting on the association between years of quitting smoking and the risk of laryngeal cancer.

The current meta-analysis had several other strengths. Firstly, the study with large sample size of 14,292 laryngeal cancer patients enabled us to quantitatively evaluate the association between cigarette smoking and laryngeal cancer risk, proving to be more forceful than any single study. Secondly, the summary association between laryngeal cancer with different types of cigarette smoking, including pack-years, frequency, and duration, were comprehensively reviewed and surveyed from a new point of perspective. Moreover, the controlled potential confounding factors within the studies could be assessed with available data from recent published studies. Sensitivity analysis and subgroup analysis were conducted to look for sources of study heterogeneity and influence of potential confounders, respectively, through identifying differences in sample size, continent, design of researches, published years of researches, scores of studies by NOS standard, and genders of participants. Furthermore, the consistent direction of the overall outcomes from subgroup analysis showed robustness of our findings. The differences in genders of participants and controls did not affect the outcomes of effect estimates obviously. The present meta-analysis illustrated that the association between cigarette smoking and laryngeal cancer risk was weaker among non-Europe and America populations than association among Europe and America populations, which might attribute to differences in the components of cigarettes and smoking habits.

Findings from this study possessed vital public health implication. The prevention of laryngeal cancer incidence remained to be an important public health problem for studies. Millions of deaths from laryngeal cancer along with other smoking-related diseases could be avoided and the burden of laryngeal cancer worldwide would be decreased by banning cigarette on a population-wide scale.

The biological mechanism that cigarette smoking led to larynx carcinogenesis remained unclear. The cancers initiated by cigarette smoke could be attributed to its various elements containing nicotine, which is the major psychoactive component, and several other toxic constituents, such as nitrosamines, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone, and polycyclic aromatic hydrocarbons. The major constituents of cigarette initiate a sequence of oncogenic events such as epigenetic changes, self-sufficiency in growth signals, evasion of apoptosis, continuous metastasis, and angiogenesis [29]. This research would give rise to new strategies for the treatment of laryngeal cancer which was caused by components of cigarettes. The capability to assess the impact of cigarette smoke on the condition of gene susceptibility with altered characters in carcinogenesis of laryngeal cancer was a promising domain for study and in favor of the better understanding of this important association. The deeper studies of correlation between risk of laryngeal cancer and similar risk factors such as tobacco chewing, smokeless tobacco, and other types of tobacco consumption and the carcinogenic effect of cigarette smoking on other subtypes of upper aerodigestive tract cancer required further investigation.

Potential limitations were present in this study. The majority of included studies were hospital-based or population-based case–control studies and thus prone to produce recall and selection bias. In addition, A few studies published in other languages for example in Polish had been searched and excluded since the limitation of inclusion criteria. Potential language bias generated when the studies not published in English or in Chinese were ruled out from the study. All available studies with disparate exposure definition were not comprised in subgroup analysis. Some studies could not eliminate former smokers from the reference group. The moderate heterogeneity within studies presenting about pack-years of cigarette smoking was mostly related to the dimension of the effect and might be interpreted by different design of studies and styles of cigarette smoking.

So far 13 eligible original studies that presented the correlation between years of quitting smoking and risk of laryngeal cancer were included. Sensitivity analysis via excluding every study was conducted to identify only seven studies for subsequent analyses. According to raw data in these seven studies, the years of quitting smoking were primarily categorized as within 15 and 16 years or more. The risk estimate in within 15 years group was generally and relatively greater than that in 16 years or more group. The characteristic of specifical risk-time development or the other points of time about the correlation between years of quitting smoking and risk of laryngeal cancer was not well understood, but certainly a critical question for future studies which would be based on lots of emerging original studies.

In summary, the results of meta-analysis presented powerful evidence that cigarette smoking was correlated with an approximately seven-time increase in risk of laryngeal cancer. The risk kept raised for 15 years after smoking cessation but dropped afterwards. This meta-analysis presented that pack-years, frequency and duration of cigarette smoking were associated with a 4.6-fold, 5.8-fold, and 4.2-fold increased risk of laryngeal cancer, respectively. The robust effect estimates were obtained through summarizing all the available studies, sensitivity analysis and subgroup analysis. The detailed analysis of association between laryngeal cancer risk and multiple risk factors such as occupational exposure, chewing of betel leaf, the other types of tobacco consumption, required future studies in more languages.