This observational longitudinal study was conducted within the CARhES cohort (Cardiovascular risk factors for health services research), which includes all adults in the Spanish Autonomous Community of Aragon (population: 1.3 million) with at least 1 cardiovascular risk factor (hypertension, diabetes, or dyslipidemia) since 2017.13

The study population comprised all CARhES participants aged ≥ 70 years who initiated statin therapy for primary prevention of CVD between 2018 and 2020. These individuals were followed for up to 2 years. Inclusion criteria were: having at least 2 statin dispensations (Anatomical Therapeutic Chemical code C10) within 180 days after initiation: at least 1 LDL-C measurement ≥ 100mg/dL (off-target) within the 90 days prior to the first statin dispensation; and at least 1 follow-up LDL-C measurement within 2 years after statin initiation. The 100mg/dL LDL-C threshold was chosen in accordance with current clinical guidelines1 for individuals at moderate cardiovascular risk, such as those in primary prevention.

Data were obtained from the CARhES cohort, whose follow-up is based on real-world data from the Aragon Health System. This is pseudonymized information on sociodemographic, clinical characteristics and healthcare services and pharmacological treatments utilization.

Socioeconomic status was inferred from annual income, which determines the level of pharmaceutical copayment. Participants were categorized as follows: income <€18 000 per annum; income ≥ €18 000 per annum; and other situations.

MACE were identified by ICD-10 codes for acute myocardial infarction (I21), nontraumatic intracranial hemorrhage, or cerebral infarction (I60-I63), recorded during hospital admissions or emergency visits.

Most sociodemographic and clinical variables were obtained from the year of the first statin dispensation, except for MACE and deaths (collected throughout the entire period). Institutionalization was considered if the participant was recorded as institutionalized at any point during follow-up. Baseline LDL-C level was based on the latest measurement within the 90 days prior to statin initiation.

The first dispensed statin was categorized according to clinical practice guidelines14 into: low-to-moderate intensity statins (simvastatin, lovastatin, pravastatin, fluvastatin, pitavastatin, atorvastatin 10/20mg, rosuvastatin 5/10mg), high-intensity statins (atorvastatin 30/40/60/80mg, rosuvastatin 15/20/30mg), and combination therapies (simvastatin+ezetimibe, pravastatin+fenofibrate, atorvastatin+ezetimibe, rosuvastatin+ezetimibe, atorvastatin+amlodipine, and atorvastatin+aspirin+ramipril).

The care level from which the initial statin prescription originated (primary care or specialty care) was also recorded.

Pharmacological burden was estimated by counting the number of unique pharmacological subgroups (third level of the Anatomical Therapeutic Chemical classification) dispensed to each participant.

Descriptive analyses of sociodemographic, clinical, and therapeutic characteristics were conducted in the total study population and stratified by sex. Continuous variables are reported as mean±standard deviation or median [interquartile range], and categorical variables as absolute frequencies and percentages. Differences between groups were assessed using the chi-square test or Fisher exact test for categorical variables, and the Student t-test for continuous variables.

Persistence was defined according to the Ascertaining Barriers to Compliance taxonomy developed by the International Society on Medication Adherence (ESPACOMP).2 In this framework, persistence refers to the time from the first dose (initiation) to the last dose taken before discontinuation.

Persistence was analyzed using the AdhereR package (version 0.8.1) in R,15 which enables transparent and reproducible medication adherence analyses using electronic health care data. Persistence to statin therapy was assessed over a maximum follow-up of 2 years. We applied the maximum gap method, whereby a patient is considered persistent as long as the interval between 2 dispensations does not exceed a predefined allowable gap. For statins, which are usually prescribed as 1 pill per day, the maximum gap was set as twice the number of pills supplied in the prior dispensation. Remaining pills from previous fills (carryover) were considered when calculating allowable gaps.

The analysis followed the Timelines-Events-Objectives-Sources (TEOS) framework16 and the ESPACOMP Medication Adherence Reporting Guidelines (EMERGE),17 both of which are consensus-based standards for transparent reporting of medication adherence research.

Differences in median persistence time among population subgroups were evaluated using the Kruskal-Wallis test and the Wilcoxon rank-sum test.

To account for competing events, we performed a competing risk analysis, treating death or MACE, whichever occurred first, as a competing risk for statin discontinuation. The cumulative incidence function (CIF) was used to estimate the probability of each event (discontinuation, MACE, or death). CIF curves were stratified by sex and age group, and by sex and statin type, with differences assessed using Gray's test.

To examine factors associated with statin discontinuation, we fitted Fine and Gray subdistribution hazard models, both crude and adjusted. The proportional subdistribution hazards assumption was verified using Schoenfeld residuals. P-values for hazard ratios (HRs) were calculated via Gray's test, and omnibus P-values were reported for categorical variables with multiple categories.

Sensitivity analyses were conducted by re-estimating the cumulative incidence functions and Fine and Gray models using an alternative allowable gap of 3 months.

Finally, the proportion of men and women achieving on-target LDL-C levels (≤ 100mg/dL) was estimated for various time windows, both overall and stratified by statin intensity. The lowest LDL-C value recorded for each patient during each time window was used. Differences in proportions were tested using the chi-square test.

All analyses were performed using R Statistical Software, version 4.1.3 (R Foundation for Statistical Computing, Austria).

A total of 4936 participants met the inclusion criteria and were included in the analyses. Their baseline characteristics are summarized in table 1. Most participants (72.1%) were aged 70-79 years. The most prevalent comorbidities were hypertension (68.9%), diabetes mellitus (29.5%), chronic kidney disease (19.0%), and depression (18.6%). Women accounted for 61.7% of the study population.

Compared with men, women had a significantly lower socioeconomic status and were more frequently institutionalized. They also had higher rates of obesity, dementia, and depression, but a lower prevalence of diabetes mellitus, ischemic heart disease, and chronic obstructive pulmonary disease.

Women exhibited significantly higher mean LDL-C levels prior to statin initiation (165 vs 152mg/dL; P<.001), a greater pharmacological burden, and were more frequently prescribed low-to-moderate intensity statins. Among women, prescriptions were more often initiated in primary care. In contrast, men had higher incidences of death and a slightly greater incidence of MACE during follow-up.

Overall, 39.9% of participants discontinued statin therapy within the 2-year follow-up period. The median time to discontinuation was 181 [89-347] days (table 2).

Among those who initiated statins and did not experience a MACE or die during follow-up (2991 women and 1822 men), 57.7% of women and 61.5% of men remained persistent over 2 years. Permanent discontinuation (no re-initiation) occurred in 12.2% of women and 11.7% of men, while 30.1% of women and 26.8% of men discontinued and later restarted therapy. Among those who reinitiated, 16.8% of women and 12.9% of men switched to a different statin.

Cumulative incidence curves revealed significant sex differences in persistence, with women remaining on treatment for a shorter duration than men (P=.003; figure 1). Participants aged ≥ 90 years and those prescribed combination statin therapies showed higher discontinuation rates than other groups, although these differences did not reach statistical significance. When stratified by sex and statin intensity, differences in persistence were evident among women: those on high-intensity statins were more likely to persist, whereas those on combination therapies showed the lowest persistence rates (P=.059; figure 2).

Figure 1.

Cumulative incidence curves of statin discontinuation and competing risks (MACE and death) (A) for the study population, (B) by sex, (C) by age group, (D) by type of statin. Discont., discontinuation; High-int-st, high-intensity statin; Low-int-st, low-intensity statin; MACE, major adverse cardiovascular event; St-comb, statin in combination.

(0.3MB).

Figure 2.

Central illustration. Sex differences in baseline characteristics, statin discontinuation incidence, and LDL-C control over time in an elderly population prescribed statins as primary prevention of cardiovascular disease. Discont., discontinuation; HDL-C, high-density lipoprotein cholesterol; High-int-st, high-intensity statin; LDL-C, low-density lipoprotein cholesterol; Low-int-st, low-intensity statin; MACE, major adverse cardiovascular event; St-comb, statins in combination.

(0.7MB).

Fine and Gray competing risk analyses showed differences between men and women (figure 3), after confirming the proportionality assumption on the subdistribution hazard scale. In women, neither the crude nor the adjusted analyses showed a statistically significant association with any of the variables studied. In men, the crude HRs for statin discontinuation were significantly lower in participants with hypertension, diabetes mellitus, heart failure, chronic obstructive pulmonary disease, and high pharmacological burden, while higher HRs were observed in those whose statins were prescribed in primary care or who had higher high-density lipoprotein cholesterol and LDL-C levels prior to statin initiation. After adjustment, only heart failure, primary care prescribing, and baseline LDL-C levels remained significantly associated with discontinuation.

Figure 3.

Crude and adjusted Fine and Gray competing risk analyses for discontinuation of statin therapy in primary prevention of cardiovascular disease in women and men older than 70 years, Aragon (Spain). Reference categories: age 70-79 years; socioeconomic level <€18 000/y; institutionalization: no; type of area: rural; comorbidities: no; type of drug: low-intensity statin; level of care: specialty care. The LDL-C and HDL-C variables were transformed so that the HRs obtained for them indicate the increase in risk as LDL-C or HDL-C increased by 10mg/dL. The adjusted models included as covariates all the other variables presented. Women Omnibus P values for age=.36; socioeconomic level=.40; type of drug=.06. Men Omnibus P values for age=.53; socioeconomic level=.45; type of drug=.66. 95%CI, 95% confidence interval; COPD, chronic obstructive pulmonary disease; HDL-C, high-density lipoprotein cholesterol; HR, hazard ratio; LDL-C, low-density lipoprotein cholesterol. *P≤.05.

(0.36MB).

In the sensitivity analyses, median times to discontinuation (table 1 of the supplementary data), based on a 3-month allowed gap, were slightly higher than those previously estimated. The results of the original Fine and Gray models were confirmed for both women (table 2 of the supplementary data) and men (table 3 of the supplementary data).

Overall, 66.1% of patients attained LDL-C targets within 6 months of statin initiation. The proportions increased over time: 68.6% at 12 months, 69.7% at 18 months, and 70.4% at 24 months.

This upward trend was observed in both women and men. However, among all time windows, women achieved LDL-C targets less frequently than men: 62.8% of women vs 71.3% of men at 6 months and 66.1% vs 77.3% at 24 months. These differences were statistically significant at all time points (P<.001; figure 2).

When statin intensity was considered, sex differences in LDL-C control persisted. Statistically significant differences across all time points were found only for low-to-moderate intensity statins (P<.0001). In addition, both women and men treated with low-moderate intensity statins had lower rates of LDL-C target attainment than those on high-intensity statins (P≤.05). For example, 6 months after initiation, 59.6% of women and 69.2% of men on low-to-moderate intensity statins were on target vs 78.3% of women and 84.3% of men on high-intensity statins.

Among those on statin combination therapy, women had a higher proportion of LDL-C target attainment than men, although the number of patients in this group was small (figure 4). Sex differences in LDL-C target attainment remained consistent after stratifying by both baseline LDL-C levels and statin intensity (data not shown).

Figure 4.

Proportion of women and men with controlled LDL-C (on target) in different window periods (months) since statin initiation, by type of statin. Only participants persistent and with at least 1 LDL-C measurement are considered in each window. Follow-up until discontinuation, MACE, or death. LDL-C, low-density lipoprotein cholesterol; MACE, major adverse cardiac events.

(0.39MB).

The main strength of the present study is its population-based nature. The findings obtained using real-world data provide a comprehensive picture of actual statin use for the primary prevention of CVD in the elderly. While there is growing evidence from observational studies on the use and effects of these drugs in clinical practice, limited evidence exists for this specific population group. It is even less common to find studies highlighting sex differences among the elderly. Data were collected from BIGAN, a platform that integrates information from the Aragon Health Service systems for secondary use in research and policy-making.28 The use of a competing risks approach enhances the validity of the results, as discontinuation is influenced by death and MACE, which are competing events that may preclude the occurrence of the event of interest.

This study also has some limitations. First, as is common in adherence studies, it was assumed that a statin was taken if it was dispensed, though this cannot be guaranteed. Nonetheless, this is currently considered a valid and reliable method for assessing persistence.29 Second, the persistence indicator calculated could yield variable and potentially non-comparable results. However, it was defined following the Ascertaining Barriers to Compliance consensus taxonomy and estimated using the AdhereR package. Consensus-based guidelines for best practices in the definition and reporting of adherence measures were followed. Furthermore, the available data did not allow us to distinguish whether discontinuation was a patient- or prescriber-driven decision.

Incorporating interval-censored survival techniques could produce more accurate estimates since both statin persistence and LDL-C target attainment were inferred from intervals. However, this approach cannot be used in conjunction with the competing risks method, which was preferred due to the short 2-month intervals, minimizing the impact of interval censoring, particularly in older populations where mortality and MACE are key competing events.

The selection and follow-up of the study population during the coronavirus disease 2019 (COVID-19) pandemic may have influenced the results. However, the 2-year follow-up period is considered sufficient to appropriately capture and assess the variables of interest. Finally, the sensitivity analyses conducted confirmed the consistency of the findings.