Long-term high-sensitivity C-reactive protein trajectories and risk of major adverse cardiovascular events in statin-treated patients: a retrospective cohort Authors: Qin Ma, Shaoping Wang, Hua Chen Published: 2026 ABSTRACT BackgroundHigh-sensitivity C-reactive protein (hs-CRP) is a key inflammatory marker in atherosclerosis; however, the association between its long-term patterns and cardiovascular prognosis under statin therapy remains incompletely understood. This study characterized hs-CRP trajectories and their prognostic value for MACE in statin users, aiming to refine risk assessment and tailored therapy.MethodsWe conducted a retrospective cohort study at the Center for Coronary Artery Disease, Beijing Anzhen Hospital, Capital Medical University (Beijing, China), using data from patients who received statin therapy for at least 12 months between January 2023 and December 2024. Eligible participants had three or more hs-CRP measurements and were free from major adverse cardiovascular events (MACEs) six months before enrollment. Latent class growth modeling was applied to categorize hs-CRP trajectories. The primary composite outcome included cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, or rehospitalization for unstable angina. Event-free survival was compared across trajectory groups using Kaplan-Meier curves with log-rank testing. Multivariable Cox proportional hazards regression evaluated the association between hs-CRP trajectories and MACE, with adjustments for age, sex, hypertension, diabetes, baseline lipids, and statin intensity.ResultsThis study included 210 patients with a median follow-up of 11.9 months. Latent class growth modeling identified three distinct hs-CRP trajectories: persistently low (45.2%, <2 mg/L), moderately fluctuating (40.0%, 2–5 mg/L), and persistently high (14.8%, ≥5 mg/L). Over follow-up, 38 patients (18.1%) experienced a MACE. Kaplan-Meier analysis revealed significantly higher cumulative MACE incidence in persistently high compared to persistently low (38.7% vs. 5.3%, P<0.001), with moderately fluctuating also showing elevated risk (25.0% vs. 5.3%, P<0.001). Following multivariable adjustment, both persistently high [hazard ratio (HR) 4.21, P=0.02] and moderately fluctuating (HR 3.46, P=0.02) hs-CRP trajectories remained independent predictors of MACE in patients receiving statins.ConclusionsStatin-treated patients with rising hs-CRP levels face elevated MACE risk. Longitudinal patterns show prognostic promise but need validation in future trials. Highlight box Key findings • Long-term high-sensitivity C-reactive protein (hs-CRP) trajectories (persistently low, moderately fluctuating, and persistently high) are independently associated with major adverse cardiovascular event (MACE) risk in statin-treated patients. • The persistently high trajectory conferred the greatest risk [adjusted hazard ratio (HR) 4.21], followed by the moderately fluctuating trajectory (adjusted HR 3.46), compared with the persistently low group. What is known and what is new? • Single-timepoint hs-CRP is a well-established biomarker for cardiovascular risk, and statins reduce both low-density lipoprotein cholesterol and hs-CRP levels. However, the prognostic value of long-term hs-CRP trajectory patterns in statin treated patients remains unclear. • This study identifies three distinct long-term hs-CRP trajectories and demonstrates that persistently high and moderately fluctuating trajectories are independently associated with increased MACE risk, even after adjusting for traditional risk factors and lipid control. The moderate fluctuation group exhibited higher within-subject variability in hs-CRP, suggesting that inflammatory instability may also have prognostic relevance. What is the implication, and what should change now? • Longitudinal hs-CRP monitoring may improve risk stratification beyond single measurements, offering a practical tool for identifying high risk patients with residual inflammation. • This trajectory based approach could inform the design of future trials evaluating anti inflammatory therapies by enabling more precise patient selection. • Prospective studies are warranted to validate these findings and determine whether intervention based on trajectory patterns improves clinical outcomes. Introduction Cardiovascular disease remains a major global cause of mortality and disability, posing a significant health threat (1). Statins, as the core drugs for regulating blood lipids and preventing atherosclerotic cardiovascular disease, have been confirmed by a large number of clinical studies to significantly reduce the risk of major adverse cardiovascular events (MACE), and are widely used in clinical practice (2). However, even with standard statin therapy, some patients still have residual cardiovascular risk, and the incidence of MACE is still high. How to accurately identify such high-risk groups and optimize management strategies is an important clinical problem (3). High-sensitivity C-reactive protein (hs-CRP) is a commonly used clinical inflammatory biomarker. Its level is closely related to the risk of cardiovascular disease, which has been included in the cardiovascular risk assessment system (4-6). It has been found in clinical practice that statins not only have lipid-lowering effects, but also reduce the level of hs-CRP to a certain extent. However, there are obvious individual differences in the change trend of hs-CRP in different patients (7). Current evidence primarily addresses single-timepoint hs-CRP and cardiovascular outcomes; however, data on long-term hs-CRP trajectory patterns and their association with MACE risk in statin-treated patients remain limited (8,9). Within this clinical context, this retrospective study examined long-term hs-CRP trajectory patterns in statin-treated patients, and explored its relationship with the risk of MACE, so as to provide practical basis for optimizing the clinical management of such patients. We also proposed the following hypothesis: there are different types of long-term hs-CRP dynamic trajectories in patients receiving standard statin therapy. MACE risk varied significantly across trajectory types. Patients with persistently high or fluctuating hs-CRP levels had a higher risk of MACE. This study is expected to identify 3–4 types of long-term hs-CRP trajectory with clinical significance, and the cumulative incidence of MACE is higher and the occurrence time is earlier in the sustained high-level group and the moderate level fluctuation group than in the sustained low-level group. Both trajectories may serve as independent predictors of MACE in this population. This approach offers a practical tool for identifying high-risk patients, and may aid in tailoring therapies to improve patient prognosis. We present this article in accordance with the STROBE reporting checklist (available at https://cdt.amegroups.com/article/view/10.21037/cdt-2026-1-0047/rc). Methods Study design This retrospective analysis utilized electronic medical records from the Center for Coronary Artery Disease, Beijing Anzhen Hospital, Capital Medical University (Beijing, China) to identify patients who had undergone consistent statin therapy for at least 12 months between January 2023 and December 2024. Data were extracted from the hospital’s electronic medical record system. The index date (time zero) for all follow-up analyses was defined as the date of enrollment, which corresponded to the completion of the 12-month statin therapy period. The first hs-CRP measurement for trajectory modelling was obtained at this enrollment visit, serving as the baseline assessment for the subsequent follow-up period. Follow-up duration was calculated from the index date to the date of first MACE occurrence or last known contact (Figure 1). Design flow chart. Ethics statement This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of the Center for Coronary Artery Disease, Beijing Anzhen Hospital, Capital Medical University (Approval No. 2022285X). Informed consent was waived for this retrospective analysis of anonymized clinical data, with all identifiers removed to ensure privacy. Inclusion and exclusion criteria Inclusion criteria (10,11): (I) age ≥18 years old; (II) regular statin therapy for at least 12 months with medication adherence ≥80% (“regular statin therapy” defined as continuous treatment for at least 12 consecutive months with no interruption exceeding 7 days; adherence assessed using the proportion of days covered (PDC) calculated from pharmacy refill records and confirmed by patient-maintained medication logs; PDC ≥80% considered adherent); (III) at least 3 times of hs-CRP detection were recorded during the study, and the detection interval was ≥3 months; (IV) and had no MACE—including cardiovascular death, nonfatal MI, nonfatal stroke, or unstable angina rehospitalization—in the six months preceding enrollment; (V) complete clinical data, laboratory examination and follow-up data. Exclusion criteria (12,13): (I) severe hepatic/renal impairment (alanine aminotransferase/aspartate aminotransferase >3× ULN, serum creatinine >177 µmol/L, or estimated glomerular filtration rate (eGFR) <30 mL/min/1.73 m2); (II) active infections (e.g., pneumonia, sepsis), autoimmune disorders (e.g., rheumatoid arthritis), and other conditions that could confound hs-CRP interpretation; (III) loss of follow-up within 6 months after enrollment or incomplete follow-up data; (IV) pregnant or lactating women; (V) recent major surgery (within 3 months), history of trauma or acute coronary syndrome; (VI) allergy to statins or inability to tolerate statin therapy. Sample size calculation The primary endpoint was the difference in MACE incidence across hs-CRP trajectory groups. Given the retrospective nature of this study, a conventional a priori sample size calculation was not feasible. Instead, we performed a post hoc power analysis based on the observed event rates to assess whether the available sample provided adequate statistical power for the primary analysis. Preliminary estimates indicated a 2-year cumulative event rate of approximately 35% in the persistently elevated trajectory group and 5% in the persistently low trajectory group. Assuming α=0.05 (two-sided), 1−β=0.90, and an allocation ratio of 1:2:2 (persistently high: moderate fluctuation: persistently low), at least 198 cases were required to consider 10% drop-out. Finally, 210 cases were included, which met the test power requirements for the primary comparison of MACE incidence across the three trajectory groups. However, we acknowledge that the sample size is limited for subgroup analyses and secondary outcomes, and the precision of effect estimates in the persistently high group is constrained by the modest number of events. Treatment options Clinical data were retrospectively obtained from medical records. All patients received guideline-recommended statin therapy. Statin intensity (low, moderate, or high) was classified per the Chinese guidelines for lipid management (2023) (14). All analyses were based on this retrospective extraction of historical intervention data. Individual statin types and specific doses were not consistently recorded in the electronic medical records and therefore could not be separately analyzed. Instead, statin therapy was categorized according to intensity (low, moderate, or high) based on the Chinese guidelines for lipid management (2023), which was used as the primary treatment covariate in the multivariate models. Evaluation indicators Data on the following evaluation indicators were retrospectively extracted and collected by consulting the hospital electronic medical record system. Baseline data Patient baseline information was retrieved from electronic medical records, including: (I) demographic characteristics: age, gender, body mass index (BMI), smoking status; (II) history of underlying diseases: hypertension, diabetes mellitus, chronic kidney disease [chronic kidney disease (CKD) stage 3–4], coronary heart disease, cerebrovascular disease, heart failure, etc.; (III) laboratory examination indicators: baseline blood lipids (total cholesterol, triglyceride, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol), liver and kidney function, fasting blood glucose, glycosylated hemoglobin; (IV) statin treatment information: drug type, treatment intensity, etc. Exposure variables hs-CRP dynamic trajectory: latent class growth modeling (LCGM) was applied to hs-CRP values measured at ≥3 time points. The optimal number of trajectories was determined by minimizing the Bayesian Information Criterion, requiring entropy >0.8, and ensuring clinical interpretability (15,16). hs-CRP measurements were obtained during routine follow-up visits as part of standard clinical care. To minimize confounding by acute inflammation, we reviewed electronic medical records and excluded any hs-CRP value associated with a documented acute infection or inflammatory condition at the time of measurement. LCGM accommodates irregular measurement intervals using time‑stamped data and full information maximum likelihood estimation, which handles missing data under a missing-at-random assumption. Main outcome measures MACE comprised the first occurrence of cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, or rehospitalization for unstable angina. Events were confirmed and adjudicated by medical records, readmission records and telephone follow-up. Secondary outcome measures Each component event of MACE: individual event rates for cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, and rehospitalization due to unstable angina were calculated. Each component event of MACE: individual event rates for cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, and rehospitalization due to unstable angina were calculated. Time of MACE: the median time of first MACE in patients of different trajectories was compared. Time of MACE: the median time of first MACE in patients of different trajectories was compared. Other inflammation-related indicators, such as baseline white blood cell count, neutrophil to lymphocyte ratio, etc. These markers were measured at baseline only. Other inflammation-related indicators, such as baseline white blood cell count, neutrophil to lymphocyte ratio, etc. These markers were measured at baseline only. Target rate of LDL-C: according to the guidelines (14), LDL-C goal achievement was defined per risk category: <1.8 mmol/L (very high-risk), <2.6 mmol/L (high-risk), and <3.4 mmol/L (low-to-moderate-risk). Statistical analysis Retrospective data analysis was conducted using SPSS version 26.0. Normally distributed continuous variables are presented as mean ± standard deviation. Intergroup comparisons were performed using one-way analysis of variance (ANOVA), with post hoc pairwise tests corrected by the Bonferroni method. Non-normally distributed variables are presented as median (IQR), analyzed with the Kruskal-Wallis H test, followed by Dunn’s post-hoc test for pairwise comparisons, with Bonferroni correction for multiple testing. Categorical data are shown as n (%) and compared via the chi-squared or Fisher’s exact test. LCGM with a censored normal distribution was applied to identify distinct hs-CRP trajectory patterns using R statistical software (version 4.3.2) with the ‘lcmm’ package (version 2.1.0). We tested models with linear, quadratic, and cubic time terms. Given the relatively short follow-up and the limited number of measurements per patient (minimum three), we prioritized parsimony to avoid overfitting. The linear model provided the best balance between goodness-of-fit and complexity, as indicated by the lowest Bayesian Information Criterion (BIC) and stable convergence; adding higher-order terms increased the BIC and did not significantly improve model fit (Lo-Mendell-Rubin test P>0.05). Accordingly, a linear function of time was assumed for each latent class. The optimal number of latent classes was determined by minimizing the BIC, requiring entropy >0.8, and ensuring clinical interpretability with each class comprising at least 10% of the cohort. To assess the robustness of the latent class solution, we conducted sensitivity analyses using alternative model specifications (varying random starts, quadratic/cubic time terms, alternative information criteria [consistent Akaike information criterion (AIC), sample-size adjusted BIC] and evaluated the stability of individual trajectory assignments across models by calculating the proportion of patients classified into the same class. Additionally, bootstrap resampling with 500 replicates was performed; in each bootstrap sample, the LCGM procedure was repeated, and the frequency of selecting the 3-class model based on the predefined criteria (BIC, entropy, LMR, class size ≥10%) was recorded. GraphPadPrism9.0 software was used to make the change trajectory diagram. Covariate selection for the multivariate Cox model was guided by established clinical and pathophysiological knowledge linking traditional risk factors to MACE in atherosclerotic populations. Age, sex, hypertension, and diabetes were included as they are well-recognized confounders in cardiovascular outcomes research. Baseline LDL-C was selected as it reflects the primary lipid target of statin therapy, and statin intensity was included to account for treatment effect variation. This set of covariates is consistent with previous literature (17); Other potential confounders such as BMI and smoking status were not included in the primary model to preserve statistical parsimony, as the limited number of MACE events (n=38) restricts the number of covariates that can be reliably adjusted. However, in a post‑hoc sensitivity analysis, we additionally adjusted for BMI and history of coronary heart disease, and the association remained robust. Results are reported as hazard ratios with 95% confidence intervals (CI). The proportional hazards assumption was tested using Schoenfeld residuals; no significant violations were detected (global P=0.28). To test the robustness of our conclusions, we performed sensitivity analyses and repeated the adjusted Cox model across subgroups. All statistical tests were two-sided, with P<0.05 considered statistically significant. Results hs-CRP trajectory identification and model selection The median follow-up time for the entire cohort was 11.9 months (IQR, 9.0–15.8 months). Follow-up durations were comparable across the three hs-CRP trajectory groups: 11.2 months (IQR, 6.8–17.0 months) in the persistently low group, 12.8 months (IQR, 11.0–15.9 months) in the moderately fluctuating group, and 11.7 months (IQR, 8.2–13.1) in the persistently high group (Kruskal-Wallis P=0.02). To identify the dynamic pattern of hs-CRP changes in patients over time, we fitted a growth mixture model with one to four latent categories. The goodness of fit indicators of each model are shown in Table 1. The AIC and other criteria continue to decline with the increase of the number of categories, but the Bayesian information criterion increases in the 4-category model (1,018.4 vs. 1,027.9), suggesting that the model may be too complex. The LMR test showed that the 4-category model did not provide significant improvement over the 3-category model (P=0.12). The three-category model had the highest entropy (0.85), indicating that its classification accuracy was the best, and the sample size proportion of all categories was more than 10%, which ensured the robustness and clinical interpretability of the results. Therefore, the 3-category model was finally adopted for subsequent analysis. Sensitivity analyses using alternative model specifications (different random starts, quadratic/cubic time terms, alternative information criteria) consistently supported the 3-class solution, with >95% of patients maintaining the same class assignment across models. Bootstrap validation further confirmed the robustness of the 3-class solution: the 3-class model was selected in 92.4% (462/500) of the bootstrap samples. The mean class proportions across bootstrap samples were consistent with the original classification (persistently low: 45.8%, moderately fluctuating: 39.2%, persistently high: 15.0%), indicating that the trajectory groups are stable and reproducible. Table 1Comparison of goodness-of-fit indices for models with different numbers of latent classesModelAICBICaBICEntropyLMR P valueBLRT P valueSmallest class probability (%)1-class1,250.41,265.11,252.3–––1002-class1,012.81,040.21,018.70.78<0.001<0.00125.23-class978.31,018.4988.20.850.003<0.00114.84-class975.11027.9988.90.820.120.138.1AIC, Akaike information criterion; aBIC, adjusted BIC; BIC, Bayesian information criterion; BLRT, Bootstrap Likelihood Ratio Test; LMR, Lo-Mendell-Rubin likelihood ratio test. Comparison of goodness-of-fit indices for models with different numbers of latent classes AIC, Akaike information criterion; aBIC, adjusted BIC; BIC, Bayesian information criterion; BLRT, Bootstrap Likelihood Ratio Test; LMR, Lo-Mendell-Rubin likelihood ratio test. Characteristics of hs-CRP trajectory The three significant trajectories ultimately identified are shown in Figure 2 and are named: persistently low level, moderately fluctuating, and persistently high level. Table 2 presents hs-CRP distribution and characteristics across the three trajectory patterns, which differed significantly in both patient composition and inflammatory burden. The proportion of patients with “persistent low level” was the highest (45.2%), and the level of hs-CRP was continuously stable at a low level (average 1.4 mg/L). The “medium level fluctuation type” accounted for 40.0%, and hs-CRP fluctuated in the medium range (mean 3.4 mg/L). The “persistently high level type” accounted for 14.8%, and the hs-CRP remained at a high level (mean 6.9 mg/L). To quantify within-subject variability, we calculated the coefficient of variation (CV) of hs-CRP. The moderately fluctuating group showed significantly higher CV than the persistently low group (median 0.32 vs. 0.18, P<0.001), supporting the presence of greater inflammatory fluctuation (see Table S1 for full data). The results clearly demonstrate significant heterogeneity in long-term inflammatory status in the context of statin therapy. Longitudinal hs-CRP trajectories. hs-CRP, high-sensitivity C-reactive protein. Table 2Characteristics and distribution of hs-CRP trajectory categoriesTrajectory classNumberMean hs-CRP (mg/L)Median hs-CRP (mg/L)Persistently low95 (45.2)1.4±0.41.4 (1.1–1.7)Moderately fluctuating84 (40.0)3.4±1.23.4 (2.5–4.3)Persistently high31 (14.8)6.9±1.26.6 (5.9–7.8)Data are presented as mean ± standard deviation, n (%) or median (IQR). The three trajectory classes represent low, moderate, and high levels of long-term inflammatory load, respectively. hs-CRP, high-sensitivity C-reactive protein; IQR, interquartile range. Characteristics and distribution of hs-CRP trajectory categories Data are presented as mean ± standard deviation, n (%) or median (IQR). The three trajectory classes represent low, moderate, and high levels of long-term inflammatory load, respectively. hs-CRP, high-sensitivity C-reactive protein; IQR, interquartile range. Comparison of baseline characteristics of patients in different trajectory groups Table 3 compares baseline characteristics across the three trajectory groups, revealing significant differences in age and BMI. Patients in the persistently high group were older and had higher BMI (P<0.05). Post‑hoc pairwise comparisons with Bonferroni correction showed that age was significantly higher in both the moderately fluctuating and persistently high groups compared with the persistently low group (adjusted P<0.05 for both), while no significant age difference was found between the moderately fluctuating and persistently high groups (adjusted P>0.05). For BMI, only the persistently high group differed significantly from the persistently low group (adjusted P=0.03); the other pairwise comparisons were not significant (adjusted P>0.05). Conventional cardiovascular risk factors (e.g., hypertension, diabetes, lipid levels) and treatment profiles (statin intensity, combination therapy) did not differ significantly among groups (P>0.05). In conclusion, although patients in the high inflammation trajectory group showed a trend toward more unfavorable baseline characteristics in age, BMI, and multiple metabolic measures, no significant differences were observed in established risk factors or treatment profiles, indicating comparability between the groups. Subsequent analyses were adjusted for age, BMI, and sex to account for potential confounding. Table 3Comparison of baseline characteristics across different hs-CRP trajectory groupsIndicatorsPersistently low (n=95)Moderately fluctuating (n=84)Persistently high (n=31)P valueDemographic characteristics Age, years63.0 (58.5–67.0)66.0 (63.0–71.0)68.0 (62.5–74.0)0.001 Male65 (68.4)54 (64.3)19 (61.3)0.72 BMI, kg/m224.5 (23.1–26.4)25.9 (23.2–27.2)26.9 (23.3–28.7)0.03 Current smoker25 (26.3)26 (31.0)12 (38.7)0.41Medical history Hypertension66 (69.5)65 (77.4)26 (83.9)0.22 Diabetes30 (31.6)36 (42.9)15 (48.4)0.14 Coronary heart disease55 (57.9)55 (65.5)22 (71.0)0.35 History of heart failure8 (8.4)11 (13.1)6 (19.4)0.21 History of cerebrovascular disease12 (12.6)15 (17.9)7 (22.6)0.37 CKD stage 3–48 (8.4)12 (14.3)6 (19.4)0.20Baseline laboratory tests Fasting blood glucose, mmol/L5.9 (5.1–6.5)6.1 (5.3–6.7)6.0 (5.7–6.8)0.10 HbA1c, %5.9 (5.5–6.4)6.1 (5.6–6.4)6.1 (5.8–6.4)0.31Baseline lipid profile Total cholesterol, mmol/L4.0 (3.6–4.5)4.1 (3.6–4.6)4.3 (3.9–4.7)0.21 Triglycerides, mmol/L1.5 (1.2–1.9)1.7 (1.1–2.0)1.8 (1.2–2.2)0.42 LDL-C, mmol/L2.4 (2.0–2.9)2.5 (2.1–2.8)2.7 (2.2–2.8)0.77 HDL-C, mmol/L1.2 (1.0–1.3)1.2 (1.0–1.3)1.0 (1.0–1.3)0.43Baseline renal function eGFR, mL/min/1.73 m286.0 (77.0–99.0)82.0 (71.5–93.0)82.0 (76.5–92.0)0.13Treatment status, n (%) High-intensity statin therapy25 (26.3)27 (32.1)13 (41.9)0.25 Combined ezetimibe treatment10 (10.5)13 (15.5)7 (22.6)0.22 Antiplatelet medication use88 (92.6)79 (94.0)30 (96.8)0.80Data are presented as n (%) or median (IQR). Continuous variables were compared using the Kruskal-Wallis H test; categorical variables using the chi-squared test. For variables with significant overall group differences (age and BMI), post-hoc pairwise comparisons were performed using Dunn’s test with Bonferroni correction. Age: persistently low vs. moderately fluctuating, adjusted P=0.04; persistently low vs. persistently high, adjusted P=0.001; moderately fluctuating vs. persistently high, adjusted P=0.33. BMI: persistently low vs. persistently high, adjusted P=0.03; persistently low vs. moderately fluctuating, adjusted P=0.32; moderately fluctuating vs. persistently high, adjusted P=0.47. BMI, body mass index; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; HbA1c, glycated hemoglobin; HDL-C, high-density lipoprotein cholesterol; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol. Comparison of baseline characteristics across different hs-CRP trajectory groups Data are presented as n (%) or median (IQR). Continuous variables were compared using the Kruskal-Wallis H test; categorical variables using the chi-squared test. For variables with significant overall group differences (age and BMI), post-hoc pairwise comparisons were performed using Dunn’s test with Bonferroni correction. Age: persistently low vs. moderately fluctuating, adjusted P=0.04; persistently low vs. persistently high, adjusted P=0.001; moderately fluctuating vs. persistently high, adjusted P=0.33. BMI: persistently low vs. persistently high, adjusted P=0.03; persistently low vs. moderately fluctuating, adjusted P=0.32; moderately fluctuating vs. persistently high, adjusted P=0.47. BMI, body mass index; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; HbA1c, glycated hemoglobin; HDL-C, high-density lipoprotein cholesterol; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol. Risk of MACE During a median follow-up of 11.9 months, 38 (18.1%) MACE occurred, with a graded increase across hs-CRP trajectory groups: 5.3% in the persistently low group, 25.0% in the moderately fluctuating group, and 38.7% in the persistently high group (P<0.001). All pairwise comparisons were significant (Dunn’s test, all P<0.05). The absolute risk difference versus the persistently low group was 19.7% for the moderate group (NNH =5) and 33.4% for the high group (NNH =3). Kaplan-Meier analysis confirmed the stratified risk pattern (log-rank P=0.005; Figure 3). Median MACE-free time was not reached in the low group, but was 17.2 months in the moderate group and 13.1 months in the high group (Table 4). Among individual MACE components, non-fatal MI differed significantly across groups (P=0.01), whereas cardiovascular death (P=0.053), non-fatal stroke (P=0.17), and rehospitalization for unstable angina (P=0.31) did not (Table 4). Kaplan-Meier curves for event-free survival from MACEs according to hs-CRP trajectory groups. The solid lines represent the cumulative event-free survival probabilities. The numbers at risk are displayed directly below the x-axis at 0, 5, 10, 15 and 20 months of follow-up. The P value was determined by the log-rank test. hs-CRP, high-sensitivity C-reactive protein; MACEs, major adverse cardiovascular events. Table 4Incidence and median time to event for primary and individual MACE across different hs-CRP trajectory groupsEventPersistently low (n=95)Moderately fluctuating (n=84)Persistently high (n=31)P valuePrimary MACE (composite endpoint)5 (5.3)21 (25.0)12 (38.7)<0.001Cardiovascular death1 (1.1)4 (4.8)3 (9.7)0.053Non-fatal myocardial infarction2 (2.1)8 (9.5)5 (16.1)0.01Non-fatal stroke1 (1.1)3 (3.6)2 (6.5)0.17Rehospitalization for unstable angina1 (1.1)4 (4.8)1 (3.2)0.31Median time to event (months)–17.213.1Log-rank P=0.005Data are presented as n (%) unless otherwise stated. Group comparisons utilized the chi-squared or Fisher’s exact test. The median time to event was analyzed using the Log-rank test. hs-CRP, high-sensitivity C-reactive protein; MACE, major adverse cardiovascular event. Incidence and median time to event for primary and individual MACE across different hs-CRP trajectory groups Data are presented as n (%) unless otherwise stated. Group comparisons utilized the chi-squared or Fisher’s exact test. The median time to event was analyzed using the Log-rank test. hs-CRP, high-sensitivity C-reactive protein; MACE, major adverse cardiovascular event. Other inflammation-related indicators and lipid control Other inflammatory markers and lipid control at the last follow-up are shown in Table 5. Systemic inflammatory burden differed significantly across the three hs-CRP trajectory groups. White blood cell count, NLR, and IL-6 all showed a graded increase from the persistently low to the persistently high group (all P<0.001). Regarding lipid control, LDL-C goal attainment (<1.4 mmol/L) showed a decreasing trend from the low to the high group (74.7% vs. 63.1% vs. 54.8%), although this difference did not reach statistical significance (P=0.07; Table 5). Table 5Comparison of other biomarkers and lipid control across different hs-CRP trajectory groupsMetricPersistently low (n=95)Moderately fluctuating (n=84)Persistently high (n=31)P valueInflammatory markers White blood cell count (×109/L)5.7±1.46.4±1.57.4±1.9<0.01 NLR2.0±0.52.3±0.82.9±1.2<0.01 Interleukin-6 (pg/mL)2.3 (1.7–3.0)3.6 (2.8–4.3)4.9 (4.4–6.2)<0.01Lipid control LDL-C goal attainment71 (74.7)53 (63.1)17 (54.8)0.07Data are presented as mean ± standard deviation, n (%) or median (IQR). The LDL-C target was <1.4 mmol/L for very high-risk patients. Continuous variables were compared using ANOVA or the Kruskal-Wallis H test; categorical variables using the chi-squared test. ANOVA, analysis of variance; hs-CRP, high-sensitivity C-reactive protein; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol; NLR, neutrophil-to-lymphocyte ratio. Comparison of other biomarkers and lipid control across different hs-CRP trajectory groups Data are presented as mean ± standard deviation, n (%) or median (IQR). The LDL-C target was <1.4 mmol/L for very high-risk patients. Continuous variables were compared using ANOVA or the Kruskal-Wallis H test; categorical variables using the chi-squared test. ANOVA, analysis of variance; hs-CRP, high-sensitivity C-reactive protein; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol; NLR, neutrophil-to-lymphocyte ratio. Multivariate Cox regression analysis affecting the risk of MACE To clarify the association between hs-CRP trajectories and MACE risk, a multivariate Cox model was fitted adjusting for age, sex, hypertension, diabetes, baseline LDL-C, and statin intensity (Table 6). hs-CRP trajectory remained significantly associated with MACE after adjustment. Relative to the persistently low trajectory, the moderately fluctuating group showed an approximately 3.5-fold higher risk (HR 3.46, 95% CI: 1.26–9.52, P=0.02), and the persistently high group showed a 4.2‑fold higher risk (HR 4.21, 95% CI: 1.32–13.44, P=0.02). None of the other covariates reached statistical significance (all P>0.05). These findings suggest that long-term hs-CRP trajectories are associated with MACE risk. Table 6Adjusted hazard ratios for MACE from multivariate Cox regressionVariableMACE events (n)HR95% CIP valuehs-CRP trajectory (reference: persistent low-level group) Moderately fluctuating213.461.26–9.520.02 Persistently high124.211.32–13.440.02Demographic and clinical characteristics Age (per 10-year increase)381.340.80–2.260.27 Male (reference: female)251.200.60–2.410.61 Hypertension (reference: normal)291.290.57–2.910.54 Diabetes (reference: normal)141.370.66–2.850.40Baseline lipid profile and treatment Baseline LDL-C (per 1 mmol/L increase)381.020.49–2.140.95 High-intensity statin therapy (reference: low/moderate)120.890.42–1.870.75The multivariate model included all listed variables. CI, confidence interval; HR, hazard ratio; hs-CRP, high-sensitivity C-reactive protein; LDL-C, low-density lipoprotein cholesterol; MACE, major adverse cardiovascular event. Adjusted hazard ratios for MACE from multivariate Cox regression The multivariate model included all listed variables. CI, confidence interval; HR, hazard ratio; hs-CRP, high-sensitivity C-reactive protein; LDL-C, low-density lipoprotein cholesterol; MACE, major adverse cardiovascular event. Sensitivity analysis To evaluate the robustness of the primary findings, we conducted several sensitivity analyses (Table 7). In the subgroup of patients with high medication adherence (≥90%), the association remained significant (persistently high: HR 3.81, 95% CI: 1.16–12.49, P=0.03), suggesting that the findings were not substantially affected by differences in adherence. After further adjustment for BMI and history of coronary heart disease, the association remained significant (persistently high: HR 4.84, 95% CI: 1.46–16.10, P=0.01), supporting the robustness of the primary results against additional confounding. Excluding patients with early MACE (within 12 months) also yielded consistent results (persistently high: HR 5.63, 95% CI: 1.05–30.15, P=0.04). In the subgroup with ≥4 hs-CRP measurements, the conclusion remained unchanged (persistently high: HR 4.73, 95% CI: 1.18–18.92, P=0.03). All sensitivity analyses produced results consistent with the primary analysis, suggesting that the association between long-term hs-CRP trajectories and MACE is robust in this cohort. Table 7Adjusted hazard ratios for the association between hs-CRP trajectories and MACE risk in various sensitivity analysesSensitivity analysis scenarioAnalyzed sample (n)MACE cases (n)Moderately fluctuatingPersistently highConsistency with primary analysisAdjusted HR (95% CI)P valueAdjusted HR (95% CI)P valuePrimary analysis (full cohort)210383.46 (1.26–9.52)0.024.21 (1.32–13.44)0.02–Additional adjustment for BMI and CHD210383.49 (1.26–9.64)0.024.84 (1.46–16.10)0.01ConsistentHigh-adherence subgroup (adherence ≥90%)185323.09 (1.08–8.85)0.043.81 (1.16–12.49)0.03ConsistentExcluding early MACE (within 12 months)200286.76 (1.50–30.43)0.015.63 (1.05–30.15)0.04ConsistentComplete data subgroup (≥4 hs-CRP measurements)178334.07 (1.17–14.19)0.034.73 (1.18–18.92)0.03ConsistentAnalyses were adjusted for age, sex, hypertension, diabetes, baseline LDL-C, and statin intensity. The additional adjustment model further included BMI and history of CHD. Statistical significance was defined as P<0.05. BMI, body mass index; CHD, coronary heart disease; CI, confidence interval; HR, hazard ratio; hs-CRP, high-sensitivity C-reactive protein; LDL-C, low-density lipoprotein cholesterol; MACE, major adverse cardiovascular event. Adjusted hazard ratios for the association between hs-CRP trajectories and MACE risk in various sensitivity analyses Analyses were adjusted for age, sex, hypertension, diabetes, baseline LDL-C, and statin intensity. The additional adjustment model further included BMI and history of CHD. Statistical significance was defined as P<0.05. BMI, body mass index; CHD, coronary heart disease; CI, confidence interval; HR, hazard ratio; hs-CRP, high-sensitivity C-reactive protein; LDL-C, low-density lipoprotein cholesterol; MACE, major adverse cardiovascular event. Discussion This retrospective cohort study of 210 patients on statin therapy characterized long-term hs-CRP trajectories and demonstrated their independent association with MACE risk. The core findings demonstrate that the long-term inflammatory status of patients, even in the context of effective statin therapy, presents significant heterogeneity and can be divided into three distinct trajectories. Compared with the “persistently low level type”, the risk of MACE in the “medium level fluctuation type” and “persistently high level type” was significantly increased. This finding suggests that long-term, dynamically monitored inflammatory burden trajectory may be a useful prognostic marker for refining risk assessment in patients receiving statin therapy, even after achieving LDL-C targets. However, given the observational design, these findings should be interpreted as associations rather than causal effects. This finding pushes the traditional perspective of cardiovascular risk management from “static measurement” to “dynamic monitoring”. In the past, as a classic inflammatory marker, the risk predictive value of hs-CRP mostly relied on single or baseline measurement, and risk stratification was carried out accordingly (18). Atherosclerosis is inherently a chronic inflammatory condition characterized by dynamic fluctuations (19). Our study applied latent class growth models that successfully captured long-term patterns of individual inflammatory activity. This confirms that the inflammatory state of patients is not always the same, but there are different evolution paths such as “continuous stability”, “level fluctuation” and “refractory increase”. The moderately fluctuating group exhibited both intermediate mean hs-CRP and greater intra‑individual variability, supporting its designation. The elevated risk in this group likely reflects the combined effect of higher average inflammation and some degree of inflammatory instability. However, the persistently high group carried the highest risk, underscoring the dominant role of absolute inflammatory burden. Whether fluctuation independently contributes to risk warrants further investigation. In particular, the identification of the “medium-level fluctuation type” trajectory reveals that the instability or variability of inflammation may serve as an independent prognostic indicator, which may be related to periodic damage of vascular endothelial function and dynamic changes in plaque stability. Therefore, long-term trajectory analysis can provide richer and more predictive panoramic information about a patient’s inflammatory activity than a single measurement that only provides an “instantaneous snapshot”. This study found that patients’ long-term hs-CRP dynamic trajectory was an independent and robust predictor of future MACE in the context of statin therapy. This conclusion is highly consistent with the current emphasis on “residual inflammation risk” in the cardiovascular field (9,20). However, the key advance of this study lies in the innovation of methodology. Previous landmark studies, landmark investigations of inflammatory pathways in cardiovascular disease have established hs-CRP as both a risk marker and a therapeutic target (4,21). A recent large collaborative analysis of more than 31,000 statin-treated patients by Ridker et al. (22) further confirmed that the residual inflammation risk represented by baseline hs-CRP had a significantly higher predictive value for future MACE, cardiovascular death, and all-cause death than the residual cholesterol risk (highest vs. the lowest hs-CRP quartile: HR 1.31; HR for cardiovascular death was 2.68). Our findings were consistent in effect size and provided stronger time order evidence with a longitudinal design. In addition, unlike the above studies, which mainly used static quantile comparison based on single or baseline hs-CRP measurement, the latent class growth model was applied to identify three dynamic trajectories of “persistently low”, “moderate fluctuation” and “persistently high” in this population. This marks an important shift in the risk assessment paradigm: from focusing on “the level of inflammation at a certain moment” to assessing “the long-term pattern and stability of inflammatory activity”. We found that patients with “medium-level volatility” had a significantly increased risk of MACE despite the fact that the average level of inflammation was not the highest, which suggests that inflammatory volatility may reflect underlying pathophysiological instability and is associated with a poorer prognosis, although causality cannot be established from this observational study, which was in line with the emerging pathophysiological hypothesis (23). From the perspective of pathophysiological mechanism, different hs-CRP trajectories may reflect distinct systemic inflammatory network status and treatment response. The persistently elevated hs-CRP pattern may reflect sustained activation of the NLRP3 inflammasome-IL-1β–IL-6 axis (24,25), or that patients have innate or acquired resistance to the anti-inflammatory effects of statins (26,27). However, given the observational nature of our data, these mechanistic interpretations remain speculative and do not establish causality. The observed associations between trajectory groups and other inflammatory markers (e.g., WBC, NLR, IL-6) support the biological validity of the trajectory classification, suggesting that hs-CRP trajectories reflect broader, coordinated innate immune activation rather than isolated biomarker abnormalities (28). Nevertheless, these findings cannot confirm that the implicated pathways are causally related to MACE risk. Experimental studies and randomized controlled trials are needed to validate the underlying mechanisms and to determine whether targeting these pathways confers cardiovascular benefit. Based on the above findings, this study has clear potential for clinical translation and provides new ideas for the precise management of cardiovascular risk. First, for risk stratification, clinical practice can consider shifting from relying on a single hs-CRP test to focusing on its long-term trends and stability. However, key questions remain before such monitoring can be recommended routinely: optimal measurement frequency, stability of trajectory classifications over time, and whether intervention based on trajectories improves outcomes. Our findings should be considered hypothesis-generating. Second, the identification of specific high-risk inflammatory trajectories could inform future trial design. For example, patients with persistently high hs-CRP might be prioritized for evaluation of anti-inflammatory therapies such as colchicine (29) or IL-6 inhibitors (30). However, this remains speculative; randomized trials using trajectory-based selection are needed to determine whether such targeted approaches enhance benefit compared to treating unselected populations. The trend toward lower LDL-C goal attainment in patients with higher inflammatory trajectories raises the possibility of an interplay between residual cholesterol risk and residual inflammatory risk. Whether the prognostic value of hs-CRP trajectories is modified by the degree of LDL-C control is an important clinical question. Although we did not have sufficient statistical power to test this interaction, future larger studies should examine whether the relationship between inflammatory trajectories and MACE differs according to LDL-C goal attainment, as this could inform more integrated treatment strategies targeting both lipid and inflammatory pathways. Of course, this study also has several limitations that need to be refined in future studies. First, the retrospective design and non-standardized nature of hs-CRP measurements introduce potential bias. Measurements were obtained during routine clinical visits rather than at prespecified intervals, and although we excluded values associated with acute inflammatory events and LCGM accommodates irregular timing, differential measurement intensity may have influenced trajectory classification. Patients with more frequent measurements may differ systematically (e.g., greater medical complexity or healthcare engagement), and medication adherence assessed by prescription records and patient logs is subject to bias. Prospective studies with standardized, regular hs-CRP assessments and objective adherence measures are needed. Second, the modest sample size—particularly the small persistently high group (n=31, 12 MACE events)—limits statistical power for subgroup analyses and examination of individual MACE components, resulting in wide confidence intervals around effect estimates. Although bootstrap validation supported the stability of the three‑class structure and sensitivity analyses adjusting for additional covariates yielded consistent results, we cannot exclude residual confounding from unmeasured factors (e.g., diet, lifestyle, frailty, subclinical atherosclerosis) or from binary classification of conditions like hypertension and diabetes that fails to capture disease severity, duration, or control quality. Larger, multicenter studies are needed to confirm these findings and enable robust subgroup analyses. Third, survival bias is inherent in the requirement for ≥12 months of statin therapy and ≥3 hs-CRP measurements, as patients with early events or loss to follow-up were excluded. This likely selects a cohort with lower baseline risk and better healthcare engagement, potentially underestimating absolute event rates and affecting generalizability. Additionally, the relatively short follow-up precludes assessment of whether the observed differences in time to MACE persist over longer periods, and the small number at risk in later follow-up limits confidence in long-term estimates. Fourth, despite multi‑modal outcome ascertainment, incomplete capture of events occurring outside our hospital system remains possible, and the modest number of events for individual MACE components (e.g., cardiovascular death, nonfatal stroke) limited our ability to detect outcome-specific associations. The significant finding for nonfatal MI should be interpreted cautiously given the lack of adjustment for multiple comparisons and the risk of type I error. Absolute risk differences and numbers needed to harm, while clinically informative, lack confidence intervals due to limited sample size and should be regarded as descriptive. Fifth, unmeasured and residual confounding cannot be excluded. Although we adjusted for key covariates, we lacked detailed data on specific statin types, doses, and their differential anti‑inflammatory effects; the numerically higher but non-significant ezetimibe use in the persistently high group may reflect either more aggressive treatment or more resistant lipid profiles. Furthermore, stage 3–4 CKD was present in 8.4–19.4% of patients and may have influenced associations, but subgroup analysis by renal function was not feasible due to sample size constraints. Sixth, the generalizability of our findings is limited by the single‑center design and highly selected population (Chinese patients from a tertiary referral center on stable statin therapy for ≥12 months with ≥3 hs-CRP measurements and no recent MACE). This cohort likely represents a more adherent, closely monitored subset, potentially limiting applicability to primary care settings, other ethnic groups, or patients with less regular follow-up. Seventh, the observational design cannot establish causality. While we interpret the findings as reflecting residual inflammatory risk, hs-CRP trajectories may serve as markers of unmeasured confounding rather than causal mediators. Factors such as detailed health behaviors, frailty, subclinical atherosclerosis burden, or residual differences in medication adherence may have contributed to both inflammatory patterns and MACE risk. Mechanistic interpretations (e.g., NLRP3 inflammasome activation) remain speculative without direct biomarker measurements, and whether targeting these pathways would confer benefit requires interventional trials. Eighth, several potentially informative analyses were not performed due to sample size limitations, including effect modification by baseline characteristics, incremental prognostic value beyond established risk scores, comparison with simpler metrics (e.g., mean hs-CRP), and predictors of trajectory membership. These questions warrant investigation in larger, adequately powered studies. Future studies should address these limitations through prospective designs with standardized hs-CRP measurements, longer follow-up, diverse populations, and detailed collection of medication data, lifestyle factors, and potential confounders. Incorporating multi-omics approaches may elucidate biological pathways underlying trajectory patterns, and randomized controlled trials using trajectory-based stratification could determine whether targeted anti‑inflammatory interventions improve outcomes in high‑risk patients. Conclusions In summary, long-term hs-CRP trajectories are independently associated with MACE risk in statin-treated patients. Dynamic hs-CRP patterns may improve risk stratification by detecting residual inflammation beyond single measurements. These findings support further research into trajectory-based strategies for selecting patients for anti‑inflammatory trials. However, causality cannot be inferred from this observational study, and prospective validation is needed. Supplementary The article’s supplementary files as Available at https://cdt.amegroups.com/article/view/10.21037/cdt-2026-1-0047/prf