3.1 Baseline characteristics according to quartiles of IBI
Table 1 outlines the characteristics of the study population, which consisted of 21,570 adult participants from the US. The demographic and clinical features of the participants were statistically represented by quartiles of IBI scores. Table 1 presents the results, which indicate statistically significant differences in demographics, including age, gender, race, education level, marital status, smoking status, alcohol consumption, household PIR, BMI, CVD, DM, hypertension, RA, stroke, and recreational activities, as well as in blood tests such as HbA1c, ALT, AST, LDL, HDL, TC, and BUN (P < 0.05). Participants who were female, non-Mexican white, highly educated, married/partnered, non-smokers, mild alcohol drinkers, with high household PIR and high BMI, and without CVD, diabetes, hypertension, or RA had higher IBI levels. As IBI scores increased, HbA1c, ALT, AST, LDL, BUN, and TC levels increased, while HDL levels decreased. Conversely, as IBI scores decreased, the prevalence of CVD, DM, hypertension, and RA increased.
3.2 Relationship between IBI and the risk of RA
Table 2 shows the results of regression analyses between IBI and RA prevalence risk obtained from different model tests. Notably, RA prevalence was positively and significantly correlated with the IBI score in all three models (P < 0.05). The IBI was converted into quartiles, and the highest quartile was associated with an increased risk of RA compared to the lowest quartile. In Model 1, participants in the highest quartile had a significantly higher odds ratio (OR) for RA prevalence compared with participants in the lowest IBI quartile (OR for Q2, 1.554 [95% CI: 1.295, 1.864]; for Q3, 2.266 [95% CI: 1.909, 2.691]; for Q4, 3.471 [95% CI: 2.948, 4.087]). In Model 2, RA prevalence was 72.3% higher in the highest quartile than in the lowest quartile (Q4, 1.723 [95% CI: 1.422, 2.087]), which was a statistically significant difference (P < 0.05). In Model 3, adjusted for all covariates, the prevalence of RA was 72.1% higher in the highest quartile group (Q4, 1.721; 95% CI: 1.418, 2.090) than in the lowest quartile group, with a significant difference (P < 0.05).
3.3 The nonlinear relationship between RA and IBI
This study analyzed the potential non-linear relationship between the risk of RA prevalence and IBI using RCS curves. The results, shown in Figure 2, indicate a significant non-linear relationship between IBI level and RA prevalence (P < 0.05). The prevalence of RA increased as the level of IBI increased, suggesting a positive correlation between RA and IBI levels.
3.4 Subgroup analyses and interactions to test the association between IBI and the risk of RA
This study used subgroup analyses to generate effect estimates for each group. Figure 3 demonstrates that the positive correlation between RA prevalence and IBI scores was statistically significant in most subgroups. However, in a portion of subgroups, the positive association was not significant. This positive association was not statistically significant (P > 0.05) among participants aged ≥65, partial race, low education level, household poverty-to-income ratio <1.3, BMI <25, partial disease (stroke, CVD, diabetes), HbA1c >6, ALT >40, AST >40, BUN >12, and HDL >60.
Interaction tests revealed that the majority of confounders in the association between IBI scores and RA did not significantly interact with this association (P > 0.05 for all interactions). Among the multiple covariates, only household poverty-to-income ratio, HbA1c, and BUN influenced this relationship.
3.5 Association between IBI and all-cause mortality in RA
The study analyzed the relationship between IBI and RA all-cause mortality using several models adjusted for different covariates. The results of the regression analyses are presented in Table 3, indicating a positive and significant correlation between all-cause mortality and IBI score in all three models (P < 0.05). When converting IBI scores to quartiles, the highest quartile was associated with increased all-cause mortality in RA compared to the lowest quartile used as the reference. Model 1, which did not adjust for any covariates, showed a significantly higher hazard ratio (HR) for all-cause mortality in participants in the highest IBI quartile compared to those in the lowest quartile (HR for Q2, 1.272 [95% CI: 0.932, 1.735]; for Q3, 1.435 [95% CI: 1.072, 1.922]; for Q4, 1.650 [95% CI: 1.250, 2.178]). In Model 2, all-cause mortality was 42.4% higher in the highest quartile than in the lowest quartile (Q4, 1.424 [95% CI: 1.043, 1.845]), which was a statistically significant difference. In model 3, after adjusting for all covariates, it was still observed that all-cause mortality was 46.4% higher in the highest quartile group (Q4, 1.464; 95% CI: 1.068, 2.007) than in the lowest quartile group. The correlation between the two was significant in all models (P < 0.05), suggesting that increased IBI scores are strongly associated with increased all-cause mortality in RA.
3.6 Non-linear relationship between IBI and all-cause mortality in RA
RCS curves were used to analyze the non-linear relationship between IBI and RA all-cause mortality. The results are presented in Figure 4, indicating a significant non-linear relationship between IBI levels and RA all-cause mortality (P < 0.05). As the IBI level increased, the RA all-cause mortality rate also increased, exhibiting an inverted J-shape.
3.7 Subgroup analyses and interaction tests for the association between IBI and all-cause mortality in RA
This study conducted subgroup analyses to estimate the impact of different subgroups on the association between IBI and all-cause mortality in RA. The results in Figure 5 show a statistically significant positive correlation between IBI score and all-cause mortality in the vast majority of subgroups. However, in a minor subgroup, the positive correlation was not significant. The positive association was not statistically significant among those with a household PIR ≥3.5, mild or heavy alcohol consumption, diabetes mellitus, ALT >40, AST >40, BUN ≤12, and HDL >60 (P > 0.05).
Interaction tests showed that the overwhelming majority of covariates did not significantly interact with each other for the positive correlation between IBI scores and RA all-cause mortality (P > 0.05). Among multiple subgroups, only race, marital status, ALT, and BUN impacted this correlation.