Triglyceride-glucose (TyG) index combined with C-reactive protein outperforms the TyG index alone in predicting stroke in arthritis patients: a nationwide prospective cohort study

Abstract

Background

See More

Arthritis patients exhibit a higher stroke risk, but effective predictive biomarkers are scarce. This research sought to examine and compare the associations between the triglyceride-glucose (TyG) index, its integration with C-reactive protein (CRP) (TyG-CRP), and stroke risk in these patients.

Methods

This research examined data from 3,419 arthritis patients participating in the China Health and Retirement Longitudinal Study (CHARLS), focusing on the occurrence of new stroke events as the main outcome. Examination of the association between the TyG index, TyG-CRP and stroke risk relied on Kaplan–Meier, Cox regression, and restricted cubic splines (RCS) analyses.

Results

During the 9-year follow-up period, 339 arthritis patients (9.9%) had their initial stroke. Stroke incidence increased steadily from the lowest to highest tertile categories as determined by both TyG index and TyG-CRP (P < 0.05). After full covariate adjustment, each 1-unit increment in TyG-CRP raised stroke risk by 18% (HR, 1.18; 95% CI, 1.01–1.39), and individuals in the top TyG-CRP tertile were 1.4 times more likely to experience a stroke versus those in the bottom tertile (HR, 1.40; 95% CI, 1.03–1.92). There was no significant link between the TyG index and stroke risk, whether it was assessed continuously or by tertiles (P > 0.05) in the fully adjusted models. TyG-CRP was significantly linearly related to stroke incidence (P-overall: 0.047; P-nonlinear: 0.725), whereas the TyG index, although linear, also demonstrated an insignificance (P-overall: 0.165; P-nonlinear: 0.557). In sensitivity analyses conducted among complete cases, TyG-CRP demonstrated borderline statistical significance for stroke risk in the model with comprehensive covariate adjustment (P = 0.058 for the continuous variable analysis; P = 0.064 for tertile-based comparisons).

Conclusion

TyG-CRP is a standalone predictor of stroke in arthritis individuals, whereas the TyG index does not significantly predict stroke risk.

Graphical Abstract

Introduction

Arthritis is a common and enduring chronic condition that manifests in several types, represented by osteoarthritis (OA) and rheumatoid arthritis (RA) [1]. Globally, arthritis is notably widespread, particularly among middle-aged people and elders [2]. In China, approximately 31.4% of middle- and older-aged adults are affected by arthritis, with prevalence rates rising with advancing age [3].

Current perspectives highlight the need to address arthritis symptoms alongside the elevated risk of major cardiovascular incidents, including stroke, among individuals suffering from arthritis [4, 5]. Stroke significantly contributes to disability and death in this population [6]. According to studies, patients with OA exhibit a 43% higher risk of stroke [7], while those with RA and ankylosing spondylitis (AS) have a 42% and 41% increased risk, respectively [8]. The increased stroke risk in arthritis patients is due to factors like limited mobility, medication side effects, inflammation, vascular dysfunction, and insulin resistance (IR) [5,6,7]. IR is a key contributor to metabolic issues, endothelial damage, and atherosclerosis, all of which are major cardiovascular risk factors [9, 10]. Thus, IR is a crucial mediator in cardiovascular events, acting as both a metabolic disruptor and vascular stressor. It is considered a reliable biomarker for predicting cardiovascular disease (CVD), particularly useful for assessing stroke risk [11, 12].

Recently, the triglyceride-glucose (TyG) index has attracted much attention as an IR indicator because of its straightforwardness, affordability, and strong link to the risk of negative cardiovascular events [13]. Furthermore, several integrated TyG indices have been developed by combining the TyG index with body mass index (BMI), waist circumference (WC), and the waist-to-height ratio (WHtR). These integrated indices have also shown predictive value for CVD; however, their predictive capability does not exceed that of the original TyG index [14]. Of note, neither the TyG index nor its derived indices incorporate inflammatory parameters, which may limit their predictive utility for diseases primarily characterized by inflammation, such as arthritis. Recently, Ruan et al. introduced an innovative approach by integrating the TyG index with C-reactive protein (CRP) (i.e., TyG-CRP), demonstrating that this combined marker provides a significant prognostic tool for anticipating survival in patients with cancer [15]. This composite marker not only reflects IR but also captures the body’s inflammatory status. Recent studies have revealed an obvious correlation between TyG-CRP and the risk of various diseases, including newly diagnosed diabetes [16] and stroke [17], within the general population. Notably, TyG alone also effectively predicts the risk of these diseases. Given that TyG-CRP accounts for inflammatory factors, its predictive efficacy for relevant diseases appears to be more pronounced in populations where inflammation is prevalent, such as those with arthritis. Despite this, current studies have not comprehensively investigated the predictive capabilities of TyG-CRP for diseases in populations characterized by inflammation, or comparatively analyzed the prediction performance of TyG-CRP and TyG.

This study seeks to analyze data from the China Health and Retirement Longitudinal Study (CHARLS) to gauge the link between TyG-CRP and the stroke risk in arthritis patients, who are more prone to experiencing strokes. Moreover, the investigation aims at comparing the prediction capabilities of the TyG index and TyG-CRP to determine if TyG-CRP can more accurately predict the stroke risk in arthritis patients.

Methods

Study population

The information used here was collected from CHARLS, an extensive, nationwide longitudinal survey targeting Chinese citizens aged ≥ 45 years. The inaugural nationwide foundational survey, known as Wave 1, took place in 2011 and involved 17,705 middle-aged individuals and older adults from 450 communities in 150 counties and districts across 28 Chinese provinces. All participants completed standardized questionnaires to gather essential information. Follow-up assessments are conducted biennially to triennially to monitor their health status. Up until the present, a total of five survey waves have been successfully carried out in the years 2011, 2013, 2015, 2018, and 2020. The data collected from each of these rounds were employed in the current study. The data collection procedures employed in CHARLS were consistent with those described in the published literature [3, 14]. Peking University’s Ethics Committee gave the green light to the CHARLS study, and all participants agreed to take part after being fully informed.

Figure 1 provides a detailed overview of how participants were selected for this study. The study commenced with the first survey wave in 2011, which had 17,705 people sign up. The analysis did not include 14,286 participants due to the established exclusion criteria. The detailed reasons for their exclusion were as follows: 6,069 individuals were not included because they lacked baseline TyG-CRP data; 1,194 participants were excluded because their baseline blood test data were obtained under non-fasting conditions; 6,859 participants were excluded either due to the absence of arthritis at baseline or the lack of baseline arthritis information; 98 participants were excluded because they either had stroke at baseline or lacked baseline stroke data; 46 participants were excluded for missing gender and age data or for being under 45 years of age; and 20 participants were excluded due to extreme values in BMI and blood pressure (BP) data. Following this comprehensive screening process, the analysis concluded with the inclusion of 3,419 patients with arthritis. Throughout the follow-up, 339 of these patients experienced a stroke, while the remaining 3,080 did not.

Fig. 1

Participant selection flowchart

Calculation of the TyG index and TyG-CRP

The TyG index is calculated by taking the natural logarithm of the product of triglyceride (TG) concentration (mg/dL) and fasting plasma glucose (FPG) level (mg/dL), divided by 2 [14]. The TyG-CRP value is derived by adding the TyG index to the product of 0.412 and the natural logarithm of CRP concentration (expressed in mg/L, denoted as Ln(CRP)), where the coefficient was derived from cancer survival models from Ruan et al. [15].

Assessment of arthritis

The assessment of arthritis relied on the self-disclosed information provided by the participants. The interview included a question asking if a doctor had ever diagnosed the individual with arthritis. Respondents who said “Yes” were identified as arthritis patients. Meanwhile, those who failed to furnish a definitive “Yes” or “No” response were omitted from the study [3].

Outcome determination

The central concern was the rate of stroke occurrences during the follow-up period. A stroke event was evaluated by posing a pivotal question to the participants: “Have you ever received a stroke diagnosis from a doctor?” The duration between the latest interview conducted and the first recorded documentation of a stroke occurrence is termed the stroke onset interval [17].

Definitions

Hypertension was identified by fulfilling one of these conditions: a systolic BP (SBP) ≥ 140 mmHg, a diastolic BP (DBP) ≥ 90 mmHg, a report of being diagnosed with hypertension by a doctor, or the administration of medication to lower blood pressure. Participants were considered diabetic if they had an FPG of 126 mg/dL or higher, an HbA1c of at least 6.5%, received a medical diagnosis of diabetes, or were on drugs for diabetes.

Covariates

The study analyzed covariates, including demographic aspects like age, gender, habits of smoking and drinking, marital status, educational background, residential location, geographic zones, and BMI; hematological parameters such as blood urea nitrogen (BUN), serum creatinine (Scr), uric acid (UA), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and platelet count; and comorbidity data, including diabetes and hypertension.

Statistical analysis

To outline the study participants’ baseline characteristics, descriptive statistics were applied. Given the data distribution, continuous variables were reported in one of two ways: either shown as the mean ± standard deviation (SD) or represented as the median along with the interquartile range (IQR). For categorical variables, they was conveyed through numerical counts alongside their respective percentages, with relevant between-group comparisons conducted via x² test. To compare continuous variables between stroke and non-stroke groups, the independent t-test served for variables exhibiting a normal distribution, and the Mann-Whitney U test served for those exhibiting an abnormal distribution. To improve the trustworthiness of the findings, this study assumed the missing values of covariates to be at random. Ten multiple imputations were conducted via chained equations (MICE package in R), with algorithm convergence evaluated via trace plots, demonstrating satisfactory convergence properties. Comparative analysis revealed that distribution of variables was similar between imputed and non-imputed datasets (Table S1), while details on the missing data for participants are provided in Table S2.

Kaplan–Meier cumulative hazard curves assisted in evaluating and comparing stroke incidence rates across tertile groups categorized by the TyG index and TyG-CRP values. The different stroke incidence rates among the tertile groups were evaluated via the log-rank test. To find out the association among the TyG index, TyG-CRP, and stroke likelihood, multivariate Cox proportional hazards regression models were applied to calculate the hazard ratio (HR) along with its corresponding 95% confidence interval (CI). Initially, the TyG index and TyG-CRP were evaluated as continuous variables to explore their associations with stroke risk. Following this, a stratification process was implemented, categorizing the participants into three distinct groups by the TyG index tertiles, where the first tertile consisted of values ≤ 8.39, the second tertile ranged from 8.40 to 8.88, and the third tertile included values > 8.88. Likewise, for the TyG-CRP, participants were stratified into three groups: tertile 1 with values ≤ 8.38, tertile 2 spanning from 8.39 to 9.06, and tertile 3 comprising values > 9.06. Three models with stepwise adjustments for confounding variables were constructed, as detailed below: Model 1 had no factors adjusted; Model 2 accounted for demographic factors such as age, gender, BMI, the use of alcohol and cigarettes, marital status, educational background, residential location, and geographic regions; Model 3 further incorporated comorbidity data, including diabetes and hypertension, along with blood parameters such as BUN, Scr, UA, TC, LDL-C, and platelet count. In addition, building upon Model 3, restricted cubic spline (RCS) regression with three specified knots was applied to investigate the possible nonlinear connections among the TyG index, TyG-CRP, and stroke risk. According to the multicollinearity test, all covariates in the fully adjusted models had a variance inflation factor (VIF) under 5 (Table S3 and Table S4), which indicates that there is no major multicollinearity present.

Subgroup analyses were conducted among predefined cohorts, categorized by age (< 60 or ≥ 60), gender, BMI (< 25 or ≥ 25), alcohol consumption, smoking status, educational attainment, marital status, residential location, geographical region, and the presence of diabetes and hypertension. To reinforce the primary findings, the study conducted multiple sensitivity analyses. Initially, to assess whether the original TyG-CRP formula was robust, the weight of CRP was recalibrated according to the specific characteristics of the study cohort. Specifically, through a multivariate Cox regression model (Model 3), the TyG index and Ln(CRP) were standardized to derive their standardized coefficients (i.e., relative weight coefficients), with values of 0.117 for TyG and 0.074 for Ln(CRP). Based on the ratio of these standardized coefficients (Ln(CRP)/TyG = 0.632), the recalibrated TyG-CRP formula was established as follows: Recalibrated TyG-CRP = TyG + 0.632 × Ln(CRP). Following this, evaluation on the association between the recalibrated TyG-CRP and stroke risk was performed within the context of Model 3. Subsequently, the robustness of TyG-CRP in predicting stroke risk was evaluated by incorporating additional adjustments in Model 3 for the utilization of antidiabetic, antihypertensive, and lipid-regulating drugs. Thereafter, a more thorough exploration of the connection between TyG-CRP and stroke incidence was undertaken by focusing solely on arthritis patients who had been followed for at least two years, thereby excluding those with shorter follow-up durations. Lastly, an analysis was conducted limited to complete cases, omitting participants with any missing covariate information. R version 4.3.3 utilized for conducting all statistical analyses, considering a two-sided P-value lower than 0.05 as statistically significant.

Results

Population characteristics

Table 1 illustrates the basic information about the participants enrolled in the study at baseline. The study included 3,419 individuals diagnosed with arthritis, with 339 and 3,080 clarified into the stroke cohort and the non-stroke cohort, respectively. The stroke group participants were generally older, showed a greater incidence of hypertension, and more commonly took antihypertensive and lipid-regulating drugs. Individuals with a stroke history showed elevated SBP, DBP, BMI, TC, TG, FPG, LDL-C, CRP, the TyG index, and TyG-CRP, accompanied by reduced high-density lipoprotein cholesterol (HDL-C). Moreover, notable geographical variations in stroke incidence were observed.

Table 1 Baseline characteristics of the arthritis population stratified by stroke status during follow-up

Association among the TyG index, TyG-CRP and stroke risk

In a follow-up period extending to nine years, 339 arthritis patients (9.9%) suffered their initial stroke. The Kaplan-Meier chart showing cumulative incidence indicated a steady rise in stroke cases across the first to third tertile groups, divided by the TyG index and TyG-CRP, with significant variations detected in both tertile groups (P < 0.05, Fig. 2).

Fig. 2

Kaplan–Meier plots depicting the cumulative risk of stroke among arthritis patients, divided by tertiles of the TyG index (A) and TyG-CRP (B)

The analysis using the Cox proportional hazards model demonstrated a notable link between stroke risk and the TyG index when considered as a continuous measure in Model 1, wherein each additional unit increment in the TyG index was linked to a 32% higher stroke risk (HR, 1.32; 95% CI, 1.14–1.53). In the model adjusted for demographic variables (Model 2), an increase of 1 unit in the TyG index was linked to a 21% greater stroke risk (HR, 1.21; 95% CI, 1.03–1.41). However, this association was rendered non-significant following further adjustments for comorbidities and blood parameters in Model 3 (HR, 1.19; 95% CI, 0.97–1.47). Similarly, when the TyG index was stratified into tertiles, the top tertile group showed a much higher risk of stroke versus the bottom tertile group in Models 1 (HR, 1.56; 95% CI, 1.19–2.03) and 2 (HR, 1.35; 95% CI, 1.02–1.78); however, such association was no longer statistically significant in Model 3 (HR, 1.17; 95% CI, 0.86–1.59). Conversely, in Model 3, treating TyG-CRP as a continuous measure maintained a statistically significant connection to stroke incidence, with each 1-unit increment correlating to an 18% greater risk of suffering a stroke (HR, 1.18; 95% CI, 1.01–1.39). Furthermore, in Model 3, the highest TyG-CRP tertile was accompanied by a greater stroke risk versus the lowest tertile (HR, 1.40; 95% CI, 1.03–1.92) (Table 2).

Table 2 Association between TyG, TyG-CRP and stroke risk in arthritis population

According to the RCS analysis, the TyG index was linearly linked to stroke risk (P-nonlinear = 0.557), but the overall association lacked statistical significance (P-overall = 0.165) (Fig. 3A). In comparison, the TyG-CRP displayed a obvious linear dose-response relationship (P-overall = 0.047; P-nonlinear = 0.725) (Fig. 3B).

Fig. 3

RCS analysis of the TyG index A, TyG-CRP B, and stroke incidence among patients with arthritis

Subgroup analysis

Analyses of subgroups and interactions were conducted specifically for the TyG-CRP to examine its association with stroke risk among various demographic groups and comorbidities. As illustrated in Fig. 4, the TyG-CRP consistently correlated with an increased incidence of stroke. No significant interactions were identified, except for a statistically meaningful interaction observed between TyG-CRP and educational level in relation to stroke risk (P = 0.046).

Fig. 4

Subgroup and interaction analyses regarding the association between TyG-CRP and stroke risk among patients with arthritis

Sensitivity analyses

Several analyses conducted here engaged in testing the sensitivity of the connection between TyG-CRP and stroke risk. Initially, the recalibrated TyG-CRP retained a statistically equivalent association with stroke risk as the original formula in Model 3 (Table S5). Subsequently, with the utilization of antidiabetic, antihypertensive, and lipid-regulating drugs further adjusted in Model 3, the analysis result that demonstrated the correlation between TyG-CRP and stroke risk conformed to the original findings (Table S6). Thereafter, this study examined above correlation among people with arthritis who had a follow-up period ≥ 2 years, and the findings conformed to the primary results (Table S7). Finally, this study conducted an analysis excluding individuals with any missing data. The results showed that treating TyG-CRP as a continuous measure resulted in a borderline significant link to stroke risk in Model 3 (P = 0.058). Similarly, when TyG-CRP was categorized into tertiles, individuals in the top TyG-CRP tertile demonstrated a borderline significant elevation in stroke risk when contrasted with those in the bottom TyG-CRP tertile in Model 3 (P = 0.064) (Table S8).

Discussion

This research marks the inaugural cohort-based investigation aimed at evaluating the associations among the TyG index, TyG-CRP, and stroke risk in arthritis patients. These findings reveal that with each 1-unit boost in TyG-CRP, there is an 18% uptick in the risk of stroke, and individuals falling into the top TyG-CRP tertile experience a risk that is 1.4-fold higher compared to those in the bottom tertile. On the other hand, the TyG index, whether measured continuously or divided into tertiles, shows no significant link to stroke risk after comprehensive covariate adjustment. Furthermore, the risk of stroke is significantly linked to TyG-CRP in a linear dose-response manner; however, this significant link is absent for the TyG index. Collectively, TyG-CRP is a standalone predictive indicator for stroke risk in individuals with arthritis, while the TyG index does not significantly predict stroke.

Among the general public, the TyG index and its composite biomarkers, such as TyG-BMI, TyG-WC, and TyG-WHtR, have been validated as effective indicators of stroke risk; however, these composites do not perform significantly better than the TyG index itself [18,19,20]. Ke et al. recently proposed the metabolic score for IR (METS-IR) as another surrogate index for IR, and such index could potentially predict CVD risk in Chinese patients with arthritis [21]. However, their Cox regression analysis only adjusted for confounders related to demographic and lifestyle factors, similar to model 2 in this study. In the present study, TyG was linked to stroke risk when adjusted for demographics, but this link disappeared after adjusting for comorbidities and blood parameters, suggesting these factors might mediate or confound the relationship. Future research should use causal inference to clarify these mechanisms. This implies that if Ke et al. had included these additional factors in their final model, they might have also found no significant link between METS-IR and CVD risk, suggesting that METS-IR might not be a dependable biomarker for forecasting CVD risk in arthritis patients. Moreover, the RCS analysis conducted in this study demonstrated that when TyG attains a value of approximately 9, the curve representing its association with stroke exhibits an attenuation phenomenon, accompanied by a marked widening of CIs. This observation may be ascribed to the scarcity of data points at the upper end of the TyG spectrum or to underlying non-linear biological effects. In contrast, although the CIs of TyG-CRP also experience a certain degree of broadening when it reaches around 10, the overall trend consistently indicates a stable and upward progression. This finding accentuates the superiority of TyG-CRP in assessing stroke risk.

In some recent studies, TyG-CRP is linked to the risk of developing a range of disorders, including newly diagnosed diabetes [16], coronary artery disease [22], and depressive disorders [23]. Nevertheless, it is crucial to emphasize that the TyG index has also been shown to have a substantial association with these diseases [24,25,26]. However, these prior studies have failed to compare the differences between the TyG index and TyG-CRP regarding their impact on disease risk. Thus, akin to the TyG index, TyG-CRP has merely been indicated to be significantly linked to disease risk. Recently, Huo et al. [17] studied the connection between TyG-CRP and the likelihood of stroke under various glycemic conditions and found a strong connection only in people with normal glucose tolerance or pre-diabetes, but not in those with diabetes. The following are proposed as the most plausible explanations for this observation. Firstly, the subgroup analysis included a relatively small number of diabetic patients, accounting for only 16.1% of the entire sample, potentially diminishing the statistical power. Secondly, the severe glycemic disturbances present in diabetic patients could have obscured or interfered with the correlation between the TyG-CRP index and stroke risk. In a similar vein, the research conducted by Huo et al. did not examine the differences in how the TyG index and TyG-CRP impact stroke risk.

The analyses of subgroups indicated no notable connection between TyG-CRP and the likelihood of stroke in diabetic or non-diabetic groups, possibly due to smaller sample sizes. However, a notable association emerged within the hypertensive group. The strength and clinical relevance of this association in hypertensive patients are strongly indicating the likelihood of synergistic vascular damage stemming from the intricate interplay between metabolic and inflammatory pathways. Hypertension is well-known for its detrimental effects on blood vessels and endothelial health, causing structural and functional alterations that predispose individuals to cardiovascular events. When combined with IR and inflammation, as reflected by the TyG-CRP index, the vascular damage appears to be exacerbated. These findings are in line with the research carried out by Tang et al. [27], which similarly found that TyG-CRP was linearly linked to the occurrence of stroke in people with hypertension. Additionally, a noteworthy connection between TyG-CRP and stroke risk was observed among males; however, no significant link was found among females, possibly due to higher estrogen levels. Estrogen plays roles in glucose and lipid metabolism and offers anti-inflammatory and antioxidant benefits, potentially helping women manage metabolic and inflammatory issues [28]. This function of estrogen could account for the feeble connection between TyG-CRP and stroke risk in women. However, due to limited sample sizes, these findings should be interpreted cautiously. The study discovered that the interaction between TyG-CRP and educational level significantly impacts stroke risk, emphasizing the need to consider education in risk assessments. This interaction could be indicative of disparities in healthcare access. Individuals with lower levels of education frequently suffer from a dearth of health knowledge, remaining oblivious to the significance of monitoring health-related indicators. Financial hardships further impede their ability to obtain tests and treatments. Additionally, those with limited education are more likely to reside in regions where healthcare facilities are in short supply. Therefore, implementing targeted interventions for individuals with both high TyG-CRP levels and low education is crucial for reducing stroke risk. Notably, TyG-CRP failed to remarkably interact with other factors, suggesting its stability as an independent stroke risk predictor.

In comparison to the TyG index, the TyG-CRP demonstrates a more robust and independent association with stroke risk in patients with arthritis. This is attributed to its integration of glucose-lipid metabolism and inflammatory processes, both of which are key pathogenic factors in arthritis and stroke [6, 7]. Although inflammation is the predominant characteristic of arthritis, disorders related to glucose-lipid metabolism also significantly contribute to its development. Research has shown that a higher arthritis occurrence in middle-aged and aged people is mainly attributed to risk factors associated with metabolic syndrome, rather than their engagement in physical activity [29]. Furthermore, the administration of lipid-lowering medications has been shown to decrease cardiovascular risks and mortality in individuals with arthritis [30]. In the context of stroke, inflammation is equally pivotal, as acute cerebral ischemia provokes immune responses that release inflammatory mediators, impacting both cerebral and systemic functions [31]. Concurrently, elevated levels of adipokines caused by metabolic disorders promote chronic low-grade inflammation and accelerate atherosclerosis, both known as major stroke risk factors [32]. More critically, there exists a complex interplay between metabolic disorders and inflammation in the pathogenesis of stroke. Metabolic disorders can induce a chronic inflammatory state that exacerbates the inflammatory response in stroke, while the acute inflammatory response elicited by stroke can further worsen the symptoms of metabolic disorders [33]. Consequently, TyG-CRP, which integrates markers of glucose-lipid metabolism and inflammation, is a highly effective indicator of stroke risk in individuals suffering from arthritis.

Strengths and limitations

This study possesses several strengths. Firstly, this research marks the inaugural endeavor to comprehensively examine and contrast the associations among the TyG index, TyG-CRP, and stroke risk in arthritis patients. Secondly, the research makes use of data from CHARLS, a cohort that represents adults in middle and older age groups across the nation. Thirdly, the research employs the multiple imputation method to appropriately address missing data in covariates, thereby enhancing statistical power. Furthermore, the robustness and reliability of the study findings are reinforced through the conduct of multiple sensitivity analyses.

This investigation is subject to several limitations. Firstly, the CHARLS database lacks granular subdivision of arthritis and stroke into specific subtypes (e.g., OA/RA, ischemic/hemorrhagic stroke). This limitation precludes subtype-specific analyses of how TyG-based biomarkers associate with stroke risk across distinct arthritis/stroke phenotypes. Critically, undifferentiated classification may introduce systematic bias: (1) The localized inflammation characteristic of OA, which is the most prevalent subtype in middle-aged/elderly cohorts, likely attenuates TyG-CRP’s predictive capacity for stroke, whereas systemic inflammation in RA may augment it; (2) Including hemorrhagic stroke, which is less associated with metabolic-inflammatory pathways, likely dilutes the true association between TyG-CRP and ischemic stroke, the subtype most mechanistically linked to IR and inflammation. Addressing this requires validation in phenotypically precise cohorts, particularly focusing on high-inflammatory arthritis subtypes and ischemic stroke. Secondly, the study only measured the baseline TyG index and TyG-CRP levels without conducting dynamic monitoring throughout the follow-up period, thereby hindering the assessment of changes in these indices and their potential impact on stroke risk. Thirdly, the reliance on self-reported diagnoses of arthritis and stroke may result in misclassification and measurement bias, potentially affecting the accuracy of the findings. These factors could cause the true relationship between TyG-related biomarkers and stroke risk to be either overestimated or underestimated. Further research is warranted to validate the findings using more objective diagnostic approaches, such as respondents presenting medical diagnosis certificates in person. Fourthly, in the sensitivity analysis excluding participants with missing data, TyG-CRP demonstrated a marginally significant association with stroke risk. This attenuation likely reflects reduced statistical power from the diminished sample size, potentially constraining the biomarker’s clinical utility. Furthermore, significant implementation barriers currently hinder TyG-CRP’s translation into clinical practice, particularly the absence of established cutoffs for stroke risk stratification, and undefined cost-effectiveness relative to conventional biomarkers. These limitations collectively underscore the necessity for definitive validation in large, independent cohorts before considering clinical adoption. Lastly, the research only involved Chinese participants who were either middle-aged or elderly, which may confer a degree of population specificity to the findings. Consequently, these findings may not hold true for other ethnic groups or nations, and caution is advised when generalizing them.

Conclusions

For arthritis patients, the baseline TyG-CRP shows a stronger and more independent link to the risk of having a stroke compared to the TyG index, suggesting that TyG-CRP could be a better predictor for stroke. Furthermore, a significant linear connection between TyG-CRP and stroke risk facilitates clinical evaluations, enabling more rapid and targeted interventions. This study also underscore the significance of accounting for population variability when selecting biomarkers to ensure precise risk prediction.