The Aragon Workers’ Health Study (AWHS) is a prospective longitudinal study designed to characterize cardiometabolic factors and assess subclinical atherosclerosis in a middle-aged Mediterranean working population. For that purpose, we analyzed data from routine annual health examinations of workers at a Spanish automobile assembly plant in Figueruelas, Zaragoza (Spain), who voluntarily agreed to participate in the study. The cohort was recruited between February 2009 and May 2010, and consisted of 5400 workers. Participants aged 40 to 55 years were additionally invited to participate in triennial imaging examinations to assess the presence of subclinical atherosclerosis and complete additional questionnaires of CV and lifestyle factors. At baseline, 587 (94.5% men) were selected for this intensive follow-up group. Most of the workers witihin the cohort were involved in manual jobs (86.5% of men and 60.5% of women). Other baseline characteristics are shown in Table 1.11 The cohort showed a high prevalence of CV risk factors and subclinical atherosclerosis but a low prevalence of clinical CV disease. Active follow-up of participants is expected to continue through 2020. Further information on the AWHS can be found in Casasnovas et al.11

We designed the present observational descriptive study to assess persistence with statin therapy in the AWHS cohort. Statin prescription data corresponding to AWHS participants was gathered from Farmasalud, the government medication consumption information system for Aragon, which is the Spanish autonomous community where the automobile factory is located. This database collects data on all prescriptions dispensed at pharmacies in Aragon through the public health care system (i.e., prescribed by either a company physician or a general practitioner from the public health care system). Prescriptions issued by private physicians, insurance companies, or in-hospital consumption, are not collected in Farmasalud. Each record in the data source corresponds to a prescription and contains the following information: an anonymous patient code, patient sex and birth date, dispensing date, Anatomical Therapeutic Chemical code of the prescribed drug, number of defined daily dose, and number of packages dispensed. Drugs were classified according to the 2015 version of the World Health Organization Anatomical Therapeutic Chemical code/Defined Daily Dose System.12 Via an encrypted code provided by AWHS researchers, we identified prescriptions for statins, both alone (Anatomical Therapeutic Chemical code C10AA) and in combination with other lipid-modifying agents (C10BA), issued between January 1, 2010 and December 31, 2014.

From the AWHS database we gathered data on CV events experienced by AWHS participants (date and event type), as well as levels of total cholesterol and other lipid fractions recorded at annual medical examinations. The low-density lipoprotein cholesterol (LDL-C) value for each patient and examination was calculated using the Friedewald formula,13 based on total cholesterol, high-density lipoprotein cholesterol, and triglyceride values.

By patient, the date of the first statin dispensing was defined as the index date. Analyses were restricted to newly-treated statin users, defined as those who had not received any statin prescription during the 12 months preceding the index date, between January 1, 2011 and December 31, 2013. Due to their low frequency in the AWHS population, women (4.2%) and men who had experienced a CV event before beginning statin therapy (secondary prevention [3.9%]) were excluded from our analyses.

Persistence was defined as continuation of treatment during a follow-up period of 1 year from the index date. Persistence was determined by measuring the time (gap) between a dispensation of a drug and the next dispensation. For this reason, participants who received only 1 statin prescription during the study period or who received 1 or more prescriptions only at the index date were excluded from the analysis. Participants were considered nonpersistent if the gap between refills was over twice the duration of the preceding prescription. While other authors have defined this gap by using a fixed number of days, we selected an alternative approach given the variable time window (1-2 months) between dispensations in the autonomous community of Aragon. The number of days of medication supplied was estimated on the basis of the number of pills and packages. In Spain, most statins are sold in packages of 28 pills, with the exception of some presentations of lovastatin, pravastatin/fenofibrate, and atorvastatin/ezetimibe, which are sold in packages of 30 pills. Of all the prescriptions dispensed to AWHS participants, only 2 corresponded to lovastatin. The usual prescribed statin dosage is 1 pill per day. Therefore, assuming that each package contained 28 daily doses, the maximum allowable gap was 56 days. The accumulation of supplies over time was not considered (i.e., only the number of pills from the most recently dispensed prescription was used to evaluate the gap). Participants were censored if the gap allowed was exceeded without purchasing a new prescription or upon reaching the end of the study period (if they had been persistent throughout the follow-up period). Nonpersistent participants were categorized as users who restarted statin therapy after a period of discontinuation, or users who simply discontinued treatment (and received no corresponding statin prescription after the end of the allowable maximum gap).

The characteristics of newly-treated statin users were recorded. Age at the index date was categorized as follows: < 50 years, 50 to 54 years, 55 to 59 years, and ≥ 60 years. Concomitant antidiabetic, antihypertensive or antithrombotic therapy were assessed by taking into account the existence of any prescription for each process, respectively, during the period between the statin index date and the discontinuation date or end of follow-up. Baseline total cholesterol and LDL-C values, recorded at the patient's last medical examination before the index date, were categorized as follows: < 100, 100 to 154, 155 to 189 and ≥ 190mg/dL for LDL-C; and < 200, 200 to 240, and > 240mg/dL for total cholesterol. These categories were established in line with the reference values provided in the European guidelines for CV disease prevention14 for individuals with low/moderate CV risk, such as those in our study population.

Persistence rates were longitudinally analyzed using the Kaplan-Meier method. Statistical differences between curves were assessed using the log-rank test. The likelihood of nonpersistence over the follow-up period was determined by multivariate analysis with Cox regression. Hazard ratios (HR) and 95% confidence intervals (95%CI) were calculated, adjusting for age group at the index date, concomitant antidiabetic, antihypertensive or antithrombotic therapy, baseline LDL-C level, and baseline total cholesterol level.

A sensitivity analysis was performed to account for the possible influence of the selected interval duration on participant classification. The proportion of persistent statin users was recalculated by applying interval durations of 1.5 and 2.5 times the number of daily doses (i.e., 42 and 70 days, respectively).

All analyses were performed using STATA version 12.1 (StataCorp, College Station, Texas, United States).

Participants in the AWHS provided written informed consent when became part of it, and the present study was approved by the Clinical Research Ethics Committee of Aragon.

Among the 5400 individuals in the AWHS, 725 new statin users fulfilled the inclusion criteria (Figure 1), with a mean (standard deviation) age of 54.7 (5.1) years. Categorization by age revealed the following breakdown: < 50 years, 13.1%; 50 to 54 years, 31.9%; 55 to 59 years, 36.8%; ≥ 60 years, 18.2%.

Figure 1.

Flowchart for study population. AWHS, Aragon Workers’ Health Study.

(0.22MB).

The median (range) baseline LDL-C value was 159.6 (58.0–294.8) mg/dL. The distribution of statin users in the different LDL-C categories was: < 100mg/dL, 26 (3.8%); 100 to 154mg/dL, 273 (39.7%); 155 to 189mg/dL, 294 (42.8%); and ≥ 190mg/dL, 94 (13.7%). The median (range) total cholesterol value was 247.0 (142.0–481.0) mg/dL, and the distribution of statin users was: < 200mg/dL, 85 (11.8%); 200-240mg/dL, 213 (29.5%); > 240mg/dL, 425 (58.8%). The median (range) of other CV risk factors was: glucose level of 98.0 (61.0–292.0) mg/dL; diastolic blood pressure of 86.0mmHg (60.0–129.0), systolic blood pressure of 128.0mmHg (90.0–220.0), and body mass index of 27.8kg/m2 (19.3–43.6).

During the corresponding follow-up period, 49.8% of statin users were being cotreated with other drugs for CV disease prevention: 36.3% with 1 additional drug, 11.7% with 2 additional drugs, and 1.8% with 3 additional drugs. Specifically, 41.4% were taking an antihypertensive, 12.3% an antithrombotic, and 11.5% an antidiabetic drug.

The most frequently prescribed statin at the index date was simvastatin (40.4% of new users), followed by atorvastatin (24.7%), rosuvastatin (24.4%), pitavastatin (5.8%), pravastatin (2.3%), simvastatin/ezetimibe (1.4%), fluvastatin (0.8%), and lovastatin (0.1%).

The 22.6% of statin prescriptions were issued by a company doctor and 77.4% by a general practitioner from the public health care system.

Of the 725 newly-treated statin users, 29.5% continued treatment for the duration of the 1-year follow-up period. The average period between the index date and treatment discontinuation was 211.3 days. According to the Kaplan-Meier analysis, half of all new statin users were treatment-persistent at 184 days after the index date, and 36.9% at 270 days after the index date. The survival curve dropped steeply at 56 days, the end of the first allowable gap (28 daily doses x 2), indicating that about 15% of new statin users discontinued treatment right after the first prescription. Among those that discontinued treatment, 57.9% restarted a statin therapy later, while 42.1% were not dispensed any other statin during the follow-up period.

Age-linked differences in persistence rates were observed (log-rank, P < .0001) (Figure 2), with persistence increasing with age. Analysis of the corresponding persistence rates for the 3 most commonly prescribed statins revealed higher rates for new simvastatin users than for rosuvastatin or atorvastatin users (log-rank P < .05).

Figure 2.

Persistence with statins for primary prevention of cardiovascular disease in male Aragon Workers’ Health Study participants, by age group.

(0.09MB).

In the bivariate analysis, the crude HR for discontinuation of statin prescription (Table 2) was significantly lower in the 2 highest age groups (55–59 years and ≥ 60 years) than in the youngest age group (< 50 years), and in users of concomitant antihypertensive treatments than in those not taking concomitant treatments for the prevention of CV disease. By contrast, higher HR values were observed for participants with baseline total cholesterol levels > 240mg/dL compared with those with levels < 200mg/dL, and in participants with baseline LDL-C levels ≥ 190mg/dL compared with those with levels < 100mg/dL. Multivariate analysis (Table 2) revealed a statistically significant HR only for statin users in the highest age group (≥ 60 years old) (HR, 0.55; 95%CI, 0.39–0.77) and in those cotreated with antihypertensive drugs (HR, 0.68; 95%CI, 0.56–0.82).

The proportion of new statin users considered treatment-persistent was 13.2% and 42.8%, applying gaps of 1.5 and 2.5 times the number of daily doses dispensed, respectively. Application of these distinct gap durations confirmed the results of the Cox regression analysis (data not shown).

The present study has several limitations. The main ones pertain to the data source used. Because the Farmasalud database only records prescriptions issued via the public health system, persistence may have been misestimated in some cases. However, it is unlikely that participants would have switched from a public health care physician to a private physician. Participants in the Aragon Workers’ Health Study have easy access to a physician, whose office is located close to their work area. Accordingly, workers usually visit the same physician in the company. Alternatively, they may visit a general practitioner within the publicly-funded health care system, in which case the prescription data would have been recorded in the same data source.

Also absent from our data source were the prescribed dose and the quantity of drug dispensed (only the number of defined daily dose was available). Given the recognized limitations of defined daily dose as a unit of measurement,18 we converted the number of pills to the daily dose based on the commercialized presentations of statins. In our opinion, this approach is valid when the recommended dose always equals a fixed number of pills.

Our data source provided information on a limited number of variables pertaining to the study participants and their treatment regimens. Information on the reasons to start treatment, family history, clinical characteristics, prescriptions for treatments other than those for the prevention of CV disease, physician characteristics, or patient attitude was not available, and would have been of interest for the identification of additional predictors of treatment persistence. Certainly, the lack of information related to the reasons to prescribe poses a determining limitation for the validity of the study, and is mainly due to the fact that the data source was not designed for this purpose.

A common limitation of analyses of drug-dispensing data is the assumption that drug purchase equates to drug consumption. Our objective was to determine rates of persistence, which may be a more appropriate indicator of treatment adherence than adherence, given that it is based on the continuity of prescription refilling behavior rather than on the number of doses available. The fact that an individual regularly attends the pharmacy to purchase their medication is a probable indicator that he is taking the prescribed medication.

In addition to the aforementioned limitations related to the data source, the variety of methodologies used to define and assess treatment persistence, as well as the lack of standardized approaches, complicates the calculation and comparison of results. Considerable variability is observed, for example, in the maximum allowable duration between dispensations. Two respective studies of Swedish8 and British17 populations measured treatment persistence in new statin users, applying a gap of 90 days from the date of the last prescription refill. The authors of the Swedish study based this decision on national regulations, which can result in a time window of up to 3 months between refills. As expected, both studies reported higher rates of statin treatment persistence (68% and 53%, respectively) than that reported here (29.5%). Although we did not initially consider a gap of 2.5 times the daily dose as the most appropriate value for evaluation of treatment persistence, the proportion of persistent users obtained using this value (42.8% of new statin users) appeared more realistic and was more in line with findings in the other populations. While our sensitivity analyses supported the robustness of the results and the gap durations applied, it should be noted that clearly defined methods and criteria for the calculation of persistence indicators are essential to ensure comparability. Several common scenarios can contribute to underestimation of statin treatment persistence. For instance, some participants can be misclassified as nonpersistent when their treatment is switched from a statin to a non-statin lipid-lowering drug. This situation is unlikely however, given that statins, alone or in combination, are recommended as the first line of treatment for hypercholesterolemia.14 Similarly, a user could purchase their first package of statins after the medical consultation but wait several days before starting treatment, thus exceeding the permitted gap and resulting in misclassification as a discontinuer.

Finally, our study population consisted only of healthy and relatively young men, which complicates extrapolation of our results to the general population of new statin users. The study population corresponds to a cohort of volunteers who are particularly aware of CV risk and undergo annual reviews and clinical analyses, as well as receiving regular and detailed health advice. We suspect that persistence with statin treatment for primary prevention of CV diseases in health care settings distinct from the AWHS may be even lower.

Nonetheless, our findings constitute an important contribution to knowledge about patterns of medication use for the prevention of CV disease, including statins. The prescription of statins in populations such as that described here remains controversial. These drugs are recommended in current clinical guidelines14 as the first-line treatment for high CV risk patients with hypercholesterolemia (nonfamilial) or combined hyperlipidemia, and are widely accessible due to their low cost. However, evidence indicates that high-potency statins are associated with an increased risk of adverse effects, including new-onset diabetes, particularly in patients at risk of developing diabetes.19 Furthermore, some authors20 have reported that pharmaceutical industry-sponsored economic evaluations of these drugs, particularly statins used for primary prevention of CV disease, generally favor the cost-effectiveness profile of the products in question.

Nevertheless, the importance of patient behavior in the management of CV risk factors is undeniable. A systematic review6 of the effects of statin treatment adherence and persistence on clinical outcomes reported that, in primary prevention, clinical benefits relating to CV events were observed after the first year of continued therapy. This underscores the importance of patient commitment and participation in CV disease control, and the need for further evaluation of the impact of persistent statin use on cholesterol levels and, consequently, on CV disease risk. Additionally, it should be taken into account that CV disease control must be managed by means of the joint control of all risk factors, not only through pharmacological therapy but also by lifestyles changes.21 Last, in those patients who do not achieve recommended targets, other therapeutic alternatives should be considered.22