Association between smokeless tobacco use and oral cavity cancer risk in women compared with men: a systematic review and meta-analysis

This systematic review and meta-analysis was performed and reported following the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) protocol [21]. The electronic databases of PubMed, Embase, and Cochrane Library were systematically searched for eligible studies from their inception up to August 2020.The following search terms were used: (“smokeless tobacco” OR “oral tobacco” OR “non burn tobacco” OR “snus” OR “gutkha” OR “naswar” OR “chew* tobacco” OR “tobacco powder” OR “tobacco tooth powder” OR “tobacco paste” OR “creamy snuff” OR “mishri” OR “masheri” OR “dip tobacco” OR “tobacco water” OR “tuibur” OR “hidakphu” OR “gul” OR “gutkha” OR “mawa” OR “khaini” OR “snuff” OR “pan masala” OR “pan masala with tobacco” OR “paan” OR “pan with tobacco” OR “zarda” OR “tambaku” OR “betel quid tobacco” OR “betel tobacco” OR “tobacco flakes” OR “tobacco leaf” OR “dried tobacco” OR “hogesoppu” OR “gnudi” OR “kadapa” OR “Mainpuri tobacco” OR “qiwam” OR “kimam” OR “dohra” OR “raw tobacco”) AND (“oral cancer” OR “oral carcinoma*” OR “oral malignant*” OR “oral tumour”). Studies reporting sex-specific relationship between SLT use and OCC risk were included. Both oral tobacco and tobacco that consumers did not smoke were included as SLT in our search strategy. No restrictions were placed on publication language and status. The references of the searched literature were also reviewed manually to further identify other eligible studies.

Two reviewers independently conducted the literature search and study selection following a standardized protocol. Discrepancies were settled by group discussion until a consensus was reached. The details regarding study inclusion criteria were as follows: (1) Participants: general population for cohort design, and OCC cases and non-cases for case-control design; (2) Exposure: SLT use; (3) Outcome: the prevalence of OCC and sex-specific effects of the relationship between SLT use and OCC risk; and (4) Study design: cohort, case-control, or case-reference studies.

OCC was defined by International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10) codes to distinguish the anatomic grouping and etiology of the disease. The studies that were included reported on cancers according with the following ICD-10 codes: C00-C06 and C09-C10, which included cancers of the lip, tongue, gum, floor of mouth, palate, cheek, vestibule of mouth, retromolar area, tonsil or oropharynx [22].

The following details of the included studies were independently extracted by two reviewers: first author, publication year, region (country in which the subject of the original study was located), study design, sample size (case/non-case), age and sex of participants, case definition, control definition, type of SLT product, confounders adjusted, matching of control, and reported sex-specific effect estimate. The Newcastle-Ottawa Scale (NOS) was used to assess the quality of observational studies, and this assessment was performed by two reviewers independently [23]. A study with 7 or more stars was considered to be of high quality, and those with 4–6 stars were regarded as moderate quality studies. Inconsistency in assessment regarding data collection and quality assessment were resolved by an additional reviewer by referring to the full text of the original article.

The association between SLT use and OCC risk was assessed using a sex-specific effect estimates with a 95% confidence interval (CI). Given the low incidence of OCC, the odds ratio (OR) was approximately equal to the relative risk (RR). A random-effects model was applied to calculate pooled RRs and 95%CIs for the relationship between SLT use and OCC risk in men and women. The female-to-male ratio of RRs (RRRs) and 95%CIs were calculated using studies that reported the direct comparisons between men and women in terms of SLT use and OCC risk. The RRRs for indirect comparisons were calculated using the studies that only reported the relationship between SLT use and OCC risk in men or women. The pooled RRRs and 95%CIs for sex-based difference in the association between SLT use and OCC risk were calculated using a random-effects model [24]. Heterogeneity across the included studies was assessed using I² and Q statistics, and I² > 50.0% or P < 0.10 was considered to indicate significant heterogeneity [25].

Sensitivity analysis was conducted to assess the stability and reliability of the meta-analysis by excluding indirect comparison results [26]. Subgroup analysis based on direct comparison results was also performed according to the type of SLT product, control definition, confounders, matching of control, and study quality. Publication bias of all study arms was calculated using funnel plot, and Egger [27] and Begg [28] test results. The P values for the pooled results were two-sided, and the inspection level was 0.05. All statistical analyses were conducted using software the STATA software (version 15.1; Stata Corporation, College Station, TX, USA).

The electronic search yielded 6132 records, and 4895 articles were retained after duplicate removal. A total of 4831 articles were excluded after reviewing the title and abstract. The remaining 64 studies were retrieved for full-text evaluations, and 2 studies were obtained by manually searching the reference lists of the 64 studies. Thereafter, 47 studies were excluded for the following reasons: effect estimates were only provided for men and women combined (n = 19), other exposures were investigated (n = 17), or insufficient data (n = 11). Finally, 19 studies were selected for the meta-analysis [29‐47] (Fig. 1).

The baseline characteristics of the included studies and participants are summarized in Table 1. The 19 included studies contained 6593 OCC cases (ranging from 84 to 1401). Fifteen studies were case-control or case-reference studies, and the remaining 4 studies were cohort studies. Thirteen studies were conducted in India, 2 in Pakistan, 2 in Sweden, 1 in the United States, and 1 study in Central and Southeast Asia. The sample size for individual studies ranged from 258 to 279,897. Seven studies had a NOS score ≥ 7 stars, and the remaining 12 studies had 5 or 6 stars.

Sixteen studies in men and 11 studies in women reported the association between SLT use and OCC risk. We noted that SLT use was associated with an increased risk of OCC in both men (RR, 2.94; 95%CI, 2.05–4.20; P < 0.001) and women (RR, 6.39; 95%C, 3.16–12.93; P < 0.001) and women had a much higher risk than did men (Fig. 2). Significant heterogeneity was observed for both studies conducted with men (I² = 92.6%; P < 0.001) and women (I² = 94.9%; P < 0.001).

A total of 10 studies directly compared the sex-based difference in OCC risk associated with SLT use, and the remaining 9 studies only reported the relationship between SLT use and OCC risk in a single-sex population. The overall pooled RRR suggested that SLT use in women was associated with an increased risk of OCC compared with that in men (RRR, 1.79; 95%CI, 1.21–2.64; P = 0.003; Fig. 3). Significant heterogeneity was found across the included studies (I² = 68.8%; P < 0.001). A significant difference was also found in the pooled RRR of indirect comparisons (RRR, 2.31; 95%CI, 1.14–4.70; P = 0.021). After excluding indirect comparison results, the conclusion was stable and not altered (RRR, 1.75; 95%CI, 1.15–2.66; P = 0.008; Fig. 4).

Subgroup analysis suggested significant sex-based difference only in individuals who received chewed smokeless products, regardless of the control definition. A pooled analysis of studies reporting on adjusted effect estimates, using matched controls, and with high quality confirmed the notably higher risk of OCC in women than that in men (Table 2).

Potential publication bias for sex-based difference in the association between SLT use and OCC risk was observed by reviewing a funnel plot (Fig. 5). However, no significant publication bias was detected through Egger (P = 0.123) or Begg test (P = 0.488).