Descriptive results
The baseline characteristics of the participants described by cancer status and sex are presented in Table 1 and Additional file 1: Table S1. This prospective cohort included a total of 5,888,034 individuals, including 2,869,265 males and 3,018,769 females. The mean age of participants at baseline was 47.28 (± 18.87) years for those without GC, 65.88 (± 10.79) years for individuals with cardia GC, and 69.27 (± 11.78) years for those with non-cardia GC. The cohort accumulated a total of 37,419,745 person-years of follow-up, with an average follow-up duration of 6.36 (± 1.98) years. Among the participants, 962,171 individuals (16.3%) were diagnosed with dental caries, 266,041 individuals (4.5%) had root canal infections, 1,101,631 individuals (18.7%) presented with mild dental inflammation, and 684,632 individuals (11.6%) were diagnosed with periodontitis. The average number of remaining teeth at baseline was 27 (± 5) (847,054 missing, 14.4%). About half of the individuals (48.7%) had a high school education, with 2.8% missing (167,665). A total of 150,303 (2.6%) participants had alcohol-related diseases, 85,233 (1.4%) had smoking-related diseases, and 96,651 (1.6%) had a family history of GC (86,348 missing, 1.5%). During the follow-up period, we identified 3993 incident GC cases (incidence rate, IR: 10.7 per 100,000 person-years), including 1241 cardia GC cases (IR: 3.3 per 100,000 person-years) and 2752 non-cardia GC cases (IR: 7.4 per 100,000 person-years) (Figs. 2 and 3).
The association between baseline dental health condition and the risk of gastric cancer in the cohort analysis. All HR and 95% CI estimates were derived from Cox models with attained age as timescale: minimally adjusted models were adjusted for sex and age at entry; fully adjusted models were adjusted for sex, age at entry, family income, education, family history of gastric cancer, smoking-related diseases, and alcohol-related diseases. Trend analyses were performed by Cochran-Armitage test. The “unknown” group was excluded from the calculation of p-trend. Abbreviations: IR, incidence rate; HR, hazard ratio; CI, confidence interval. *: p < 0.05; **: p < 0.01; ***: p < 0.001
The association between baseline dental health condition and the risk of cardia and non-cardia gastric cancer in the cohort analysis. All HR and 95% CI estimates were derived from Cox models with attained age as timescale, adjusted for sex, age at entry, family income, education, family history of gastric cancer, smoking-related diseases, and alcohol-related diseases. Trend analyses were performed by Cochran-Armitage test. The “unknown” group was excluded from the calculation of p-trend. Abbreviations: IR, incidence rate; HR, hazard ratio; CI, confidence interval. *: p < 0.05; **: p < 0.01; ***: p < 0.001
Associations between dental health and GC and its subtypes
As shown in Figs. 2 and 3, our main analyses revealed a positive association between dental health condition and the risk of GC, and an inverse association between the number of remaining teeth and the risk of GC, although the effect size was slightly attenuated in the fully adjusted models.
In the fully adjusted models, associations between odontogenic inflammation and non-cardia GC were insignificant despite hazardous point estimate. The same pattern was observed for cardia GC except for periodontitis. Combining two subtypes did not change the direction or magnitude of the association. When compared to the healthy group, individuals suffering from periodontitis had an 11% (95% CI: 1% to 21%) and 25% (95% CI: 7% to 46%) increased risk of GC and cardia GC, respectively.
The near-linear, negative exposure-response curves suggest decreasing trends in the risk of GC or its two subtypes as the number of remaining teeth at baseline increased (Fig. 4). When comparing individuals with more than 27 remaining teeth (the reference group), those with 24 to 27, 20 to 24, 14 to 20, and 14 or fewer remaining teeth had increased risks of GC by 4% (95% CI: − 5% to 15%), 26% (95% CI: 13% to 40%), 46% (95% CI: 30% to 64%), and 55% (95% CI: 36% to 76%), respectively. Similarly, the risks of cardia GC were increased by 6% (95% CI: − 11% to 25%), 38% (95% CI: 15% to 66%), 74% (95% CI: 41% to 114%), and 70% (95% CI: 34% to 117%) in these groups. For non-cardia GC, the risks were increased by 4% (95% CI: − 8% to 17%), 21% (95% CI: 7% to 38%), 36% (95% CI: 18% to 56%), and 49% (95% CI: 28% to 74%), respectively.
Exposure-response curve for the association of remaining tooth number at baseline with the risk of gastric cancer and its anatomical subtypes among individuals in the Swedish Dental Health Register, 2009–2016. A Total gastric cancer; B cardia gastric cancer; C non-cardia gastric cancer. The dark blue solid line represents the point estimates and the black dash lines indicate corresponding 95% CIs, which were derived from Cox models with attained age as timescale, adjusted for sex, age at entry, family income, education, family history of gastric cancer, smoking-related diseases, and alcohol-related diseases
There was no significant interaction between dental inflammatory conditions and the number of remaining teeth on the risk of GC or its subtypes (p for interaction: 0.379 for total GC, 0.561 for cardia GC, and 0.345 for non-cardia GC). As shown in Additional file 1: Table S2, compared to individuals who had no inflammation and more than 27 remaining teeth, other groups generally showed increased risks of GC and its subtypes. For example, those with mild or severe dental inflammation and ≤ 14 remaining teeth had higher risks for total GC, cardia GC, and non-cardia GC, with HRs (95% CIs) of 1.68 (1.38, 2.03), 1.80 (1.25, 2.59), and 1.62 (1.29, 2.04), respectively.
Sibling comparison
For the sibling-controlled study, we matched 4022 unaffected full-sibling controls for 1987 incident GC cases (Table 2). Consistent with the cohort analysis, individuals with dental health problem had a higher risk of developing GC. While some ORs were either insignificant or marginally significant due to limited statistical power, the overall trends were clear. For example, the ORs (95% CI) for GC were 1.25 (1.05, 1.49) for individuals with dental caries, compared to the healthy group. Additionally, we observed a gradual increase in GC risk with a decreasing number of teeth at baseline. Individuals with 20 to 24, 14 to 20, and ≤ 14 remaining teeth had increased risks of 28% (95% CI: 6% to 55%), 43% (95% CI: 13% to 81%), and 47% (95% CI: 10% to 95%) for GC compared to the reference group, which aligns with the estimates from the cohort analyses.
Subgroup analysis
The associations between dental health and GC, as well as its main subtypes, among different sex and age subgroups are shown in Additional file 1: Figs. S2–S5. Consistent with our main analyses, significant positive associations or trends were observed in most subgroups. For example, compared to the healthy group, the HRs (95% CI) for GC and cardia GC risk in the periodontitis group were 1.17 (1.05, 1.31) and 1.29 (1.08, 1.54) for male participants, and 1.25 (1.11, 1.41) and 1.44 (1.19, 1.74) for individuals under 70 years old. Additionally, compared to participants with more than 27 remaining teeth, those with ≤ 14 remaining teeth at baseline had HRs (95% CI) of 1.60 (1.36, 1.88), 1.58 (1.20, 2.09), and 1.59 (1.30, 1.94) for GC, cardia GC, and non-cardia GC, respectively, in male participants, and 1.75 (1.41, 2.17), 1.90 (1.35, 2.69), and 1.67 (1.27, 2.19) in participants under 70 years old. However, the associations between dental health condition and non-cardia GC in males, females, and those over 70 years old, as well as the associations between dental health condition and total GC in females and those over 70, were only marginally significant.
Sensitivity analysis
Our sensitivity analysis, which excluded incident cases and follow-up data from the first 2 years (Additional file 1: Fig. S6) as well as participants with missing covariate data (Additional file 1: Fig. S7), further supported the significant associations between poor dental health, fewer remaining teeth at baseline, and increased risks of GC and its subtypes. Despite reduced stability in some association estimates, all these findings were consistent with our primary results.