This study utilized a continuous enrollment approach, recruiting voluntary participants who met eligibility criteria in the neurology department of a tertiary grade A comprehensive hospital in Fujian Province, China, from January 2021 to May 2023. The total study duration spanned from January 2021 to May 2024, comprising two phases: (1) the enrollment phase (January 2021 to May 2023) and (2) a 12-month follow-up period for all participants (until May 2024). Data analysis was conducted following the conclusion of the follow-up in May 2024.

Inclusion criteria were as follows: (1) Age ≥ 18 years. (2) Diagnosis confirmed according to the Guidelines for the Diagnosis and Treatment of Carotid Artery Stenosis (1). (3) Underwent carotid artery stenting. (4) For patients with a history of stroke, no significant residual sequelae, with a modified Rankin Scale (mRS) score of 0–2 (16). Exclusion criteria encompassed: (1) Patients with cardioembolic sources and intracranial atherosclerosis (ICAS). (2) Patients with impaired consciousness and communication disorders who cannot cooperate with the investigation. (3) Patients with severe mental illness and psychological disorders. (4) Patients with significant cardiac, pulmonary, or renal dysfunction, severe liver damage, or malignant tumors. Dropout criteria were as follows: (1) Patients lost to follow-up. (2) Patients who withdrew from the study, including those unreachable after two consecutive contact attempts, with a 1-day interval between attempts.

This study is a single-center investigation that strictly adhered to standardized protocols for patient screening, data collection, and evaluation, employing uniform inclusion and exclusion criteria to minimize data variability. Comprehensive data were gathered, including demographic characteristics, comorbidities, the number of stents implanted, and follow-up assessments. This meticulous data collection facilitates precise adjustment for confounding factors during analysis. Personnel involved in data collection and evaluation underwent rigorous training and calibration, complemented by regular audits to ensure measurement consistency and accuracy. Furthermore, ongoing monitoring and quality control checks were implemented to uphold adherence to the study protocol and to promptly identify and rectify any deviations, thereby preserving the integrity of the collected data.

This study initially enrolled 250 participants, with 247 completing the baseline survey. Of these, 242 patients completed the T1 assessment, 236 completed the T2 assessment, and 209 completed the T3 assessment. Individuals lost to follow-up (5 at T1, 4 at T2, and 24 at T3) and those who died (2 at T2 and 3 at T3) were excluded from the analysis. Consequently, 209 participants with complete MMAS-8 data at all three time points were included in the GBTM analysis. Of these participants, 171 were male (81.82%) and 38 were female (18.18%); the average age was 66.8 ± 9.2 years. In terms of stent implantation, 110 cases (52.63%) had one CAS stent implanted, 61 cases (29.19%) had two, and 38 cases (18.18%) had three. Over 12-month follow-up, 20 cases of recurrent stroke were observed. Refer to Figure 1 and Table 1 for detailed data.

Medication adherence scores at T1, T2, and T3 were 6.33 ± 1.43, 5.06 ± 1.31, and 4.81 ± 1.47, respectively, Medication adherence was analyzed as a continuous variable using GBTM with a censored normal distribution. Participants with fewer than three adherence measurements were excluded. Models with 1 to 4 groups were initially fitted using linear, quadratic, and cubic terms. The 4-group model was discarded as one trajectory group comprised less than 5% of the sample, resulting in the selection of a 3-group model. Based on trajectory plots and fit indices, the final model included three quadratic trajectories. See Supplementary Tables S1, S2.

This model categorizes patients into three distinct trajectories: Group 1, the “Persistently low level decline” cohort (31.10%), characterized by consistently low medication adherence throughout the study period, with a notable decline at 6 months, followed by a gradual decrease. Group 2, the “Medium level decline” cohort (31.58%), includes patients with initially moderate adherence, experiencing a sharp decline at 6 months before stabilizing. Group 3, the “High level decline” cohort (37.32%), represents patients with initially high adherence, undergoing a rapid decline at six months, followed by stabilization. See Figure 2.

Multivariable regression analysis indicated monthly household income, family medication supervision, regular follow-up visits, and medication-related beliefs significantly influenced the trajectories of patients with CAS. Specifically, patients with a monthly household income exceeding 6,000 yuan were 17.562 times and 25.277 times more likely to transition to the medium level and high level decline groups, respectively, compared to those with an income below 3,000 yuan. Additionally, Patients without family supervision for medication adherence were less likely to transition to the high decline group, with an odds ratio of 0.07. respectively, compared to those with family supervision. Regular follow-up visits were associated with increased odds of transitioning to the medium level and high level decline groups, with odds ratios of 30.405 and 26.123, respectively, compared to patients with irregular follow-ups. Moreover, higher scores in medication beliefs were significantly associated with an increased likelihood of transitioning to both the medium level and high level decline groups, as shown in Tables 1, 2.

The 12-month stroke recurrence rate following CAS was 9.57%. Kaplan-Meier survival analysis indicated significant differences in stroke recurrence rates among the three adherence trajectory groups over time (Log-rank and Breslow test, both P < 0.001). Notably, with extended follow-up, the persistently low level decline group exhibited earlier recurrences and the most rapid increase in recurrence rates, indicating a higher risk compared to the high level decline group. In Cox proportional hazards models, both Model 1 and Model 2 analyses showed that patients in the high level decline group had a significantly lower risk of stroke recurrence compared to those in the persistently low level decline group (P < 0.05). Conversely, Model 3 revealed no significant association (P > 0.05). Detailed results are presented in Table 3 and Figure 3.

Historical studies on medication adherence trajectories indicate significant variability across diseases and medications, with trajectory groups ranging from three to seven (11). For example, a study on statin adherence identified six trajectories: 21.1% of patients exhibited perfect or near-perfect adherence, while 19.5% were nearly completely nonadherent. The remaining 59.4% were classified as moderate adherers, with four subgroups: rapid decline, slow decline, occasional use, and brief gaps. Importantly, these statin adherence trajectories have been shown to predict future cardiovascular events (24). A cohort study on hypertension identified six medication adherence trajectories: 40% of patients were adherent, 10% experienced an early drop-off followed by a rebound to nearly full adherence, 10% exhibited partial drop-off, 14% showed gradual drop-off, 8% demonstrated rapid drop-off, and 18% had immediate drop-off. This underscores that medication adherence declines with an increase in the number of medications and treatment complexity (25). In a Chinese cohort of coronary artery disease patients (26), four adherence trajectories were identified: persistent adherence (39.9%), increasing then decreasing adherence (23.1%), increasing adherence (23.4%), and nonadherence (13.6%). Several factors may account for these discrepancies, including disease characteristics, treatment regimens, patient populations, and healthcare system differences.

Compared with more prevalent conditions such as hypertension and coronary artery disease, CAS is less common, and patients receive limited health education, which may contribute to lower medication awareness and a progressive decline in adherence over time. Additionally, differences in assessment tools may significantly impact study findings. This study utilized the MMAS-8 scale (27), which evaluates both medication-taking behavior and attitudes, and has been shown to perform well in identifying low adherence (28), whereas approximately two-thirds of adherence trajectory studies rely on proportion of days covered (PDC) metrics derived from pharmacy records (11). However, PDC is highly dependent on healthcare system structures, and in China, where patients do not consistently obtain medications from a single pharmacy, its reliability is limited. Furthermore, this study assessed adherence at 3, 6, and 12 months, which, while capturing overall trends, may not fully capture short-term fluctuations in adherence trajectories.

Medication adherence trajectories are shaped by multiple factors. Regular follow-up has been shown to reduce non-adherence (29), yet studies examining its impact on adherence trajectories are limited. Our analysis demonstrates that patients receiving regular follow-ups are 26.123 times more likely to fall into the high-level decline group, emphasizing the role of follow-up as a key predictor of adherence. Furthermore, we identify the lack of family supervision as a novel independent factor influencing adherence trajectories. The demographic characteristics of our cohort, including an average age of 66.8 ± 9.2 years and a prevalence of chronic comorbidities (82.8%), highlight the essential role of caregivers in medication management. Additionally, lower household income and negative medication beliefs independently affect adherence trajectories, aligning with findings in the existing literature (30, 31).

Independent factors influencing medication adherence trajectories after CAS provide critical opportunities for intervention in this patient population. First, evaluating financial barriers and medication beliefs shortly after discharge is practical and facilitates early identification of high-risk individuals. Second, incorporating structured caregiver involvement in medication management—through initial education and ongoing feedback—strengthens adherence oversight. Fourth, the incidence of mortality and loss to follow-up during the study led to a scarcity of data points, necessitating their exclusion from the analysis and obstructing a thorough evaluation of postoperative outcomes in CAS patients. Lastly, the single-center design might restrict the generalizability of these findings, highlighting the need for further research to address these limitations.

This study investigates the association between medication adherence trajectories post-CAS and stroke recurrence, underscoring the importance of the high level decline group as a potential protective factor. The protective effect may stem from high medication adherence postoperatively, which is critical for thwarting stroke recurrence by inhibiting platelet aggregation and minimizing thrombotic risk (32, 33). While confounding variables hinder the establishment of a direct causal relationship, the findings endorse adherence trajectories as essential tools for clinical management. Prior research has established that low medication adherence is an independent predictor of adverse outcomes in chronic hepatitis B (CHB) and acute coronary syndrome (ACS) patients (34, 35). This study further affirms its clinical significance in preventing stroke recurrence following CAS.

The clinical implications of this study are as follows: clinicians should adopt early risk prediction and stratification approaches in their management strategies, informed by medication adherence trajectories post-CAS. First, medication adherence at 3 months post-procedure should be utilized to forecast subsequent trajectory patterns and stratify risk, thereby identifying high-risk populations based on influencing factors. Second, for patients in the persistently low adherence decline group, intensive interventions must be implemented to address medication-related concerns. This may include simplifying prescriptions based on financial constraints, providing low-cost alternatives, increasing follow-up frequency, incorporating electronic monitoring systems, and engaging caregivers in the intervention process. For the moderate decline group, it is essential to enhance positive behaviors through adherence reward programs and identify barriers. Patients in the high-level decline group should receive ongoing support, with a focus on reinforcing necessary beliefs to mitigate burnout. Finally, intensifying interventions within the first 6 months postoperatively is critical to preventing declines in adherence.

This study has several limitations. First, we assessed medication adherence among CAS patients at only three time points, which precluded a detailed understanding of the trajectory curves. Second, future studies should increase the sample size and incorporate potential mediating variables to clarify the pathways by which medication adherence trajectories affect stroke recurrence. This will not only validate our findings but also enhance their relevance for broader clinical applications. Third, while numerous factors influence medication adherence, this study did not account for personality traits or medication-related variables, resulting in a lack of more nuanced insights. Fourth, the incidence of mortality and loss to follow-up during the study contributed to incomplete data points, consequently omitting these cases from trajectory analysis and restricting a comprehensive evaluation of prognosis in CAS patients. Lastly, the single-center design of this study may constrain the generalizability of the findings, highlighting the need for additional research to address these limitations.