Data were obtained from Intego, a general-practice-based morbidity registration network at the Department of Public Health and Primary Care of the University of Leuven, Belgium [37]. The Intego procedures were approved by the ethical review board of the Medical School of the University of Leuven (N° ML 1723) and by the Belgian Privacy Commission (no SCSZG/13/079). In 2013, 111 general practitioners (GPs), all using the medical software Medidoc®, collaborated in the Intego project. They worked in 48 practices evenly spread over Flanders, the Northern part of Belgium. GPs applied for inclusion in the registry. Before acceptance of their data, registration performance was audited using a number of algorithms that compared their results with those of all other applicants. Only the data of the practices with an optimal registration performance were included in the database. The Intego GPs prospectively and routinely registered all new diagnoses together with new drug prescriptions, laboratory test results and some background information (including gender and year of birth), using computer-generated keywords internally linked to codes. New data were coded and collected from the GPs’ personal computers with specially framed extraction software and entered into a central database. Registered data were continuously updated and historically accumulated for each patient. New diagnoses were classified according to a very detailed thesaurus automatically linked to the International Classification of Primary Care (ICPC-2). Drugs were classified according to the WHO’s Anatomical Therapeutic Chemical (ATC) classification system.

The current study is a retrospective cohort study that used Intego data from January 1st 1999 to December 31st 2013. In every yearly contact group (patients in contact with their GP for either reason during a year) between 1999 and 2013, we selected all patients aged 50 years or older with a history of CVD (myocardial infarction (MI) (K75), stroke (K90), transient ischemic attack (TIA) (K89), ischemic heart disease (IHD) with and without angina (K74 and K76, respectively) and peripheral arterial disease (PAD) (K92)). Recurrent CVD was defined as an incident case of MI, stroke or TIA in all selected patients, or an incident diagnosis of IHD or PAD in patients without a previous diagnosis of IHD or PAD. MI and stroke were considered as major events. TIA, IHD and PAD were considered as minor events. The prevalence of statin (ATC code C10AA) use (at least 2 prescriptions in the selected year) was registered for each year.

To analyse time trends in age-standardized rates between 1999 and 2013, a joinpoint regression analysis was performed [40]. A joinpoint is a point in the trend curve where a statistically significant change in trend over time is observed. A minimum number of 3 observations from a joinpoint to either end of the data, and a minimum number of 4 observations between two joinpoints were required. The age-standardized rates were computed taken the Flemish population in Belgium as the standard population, using 10-year age groups until 79 years, and 80 years and older as the last age group for standardization. The reference year for the standard population was 1999. From the joinpoint regression model, the annual percentage change (APC) and the average annual percentage change (AAPC) were extracted. APC is calculated for each significant trend from a piecewise log-linear model on the logarithm of the age-standardised rate versus the year. AAPC represents the average of APC estimates per significant trend weighted by the corresponding trend length (number of years in the trend). The trend analysis using the joinpoint regression model was performed using the SEER*Stat software (Joinpoint Trend Analysis software from the Surveillance Research Program of the US National Cancer Institute (available at http://​surveillance.​cancer.​gov/​joinpoint)).

Continuous data were summarized using median [P₂₅; P₇₅], and categorical data using proportions. Factors associated with statin use in patients in secondary prevention in 2013 were assessed using mixed effects logistic regression, which belongs to generalized linear mixed models (GLMM). Each GP practice was treated as a random effect to explore the variability between GP practices in addition to individual patient observations themselves, by age group (50–59 years, 60–79 years, 80+ years). The log-likelihood was estimated using the Laplace approximation. All variables in univariate analysis were candidate for the multivariable model. A stepwise approach was then used to select the best multivariable model. The goodness of fit of the model was assessed using the Akaike information criteria (AIC), Hosmer-Lemeshow test, c-index of the receiver operating characteristic (ROC) curve and the Somers’ D_xy rank correlation. The c-index assesses the predictive performance of the model and the Somers’ D_xy is an estimate of the rank correlation of the observed binary outcome and the predicted probabilities. In addition, at each step, the maximum variance inflation factor value associated with each parameter in the model was required to be less than 10 for the model to be chosen. A two-sided p-value < 0.05 was considered statistically significant. Statistical analyses of the factors associated with statin use in patients in secondary prevention were performed using R Software Version 3.0.3 (Free Software Foundation Inc., Boston, MA, USA) (lme4, pgirmess, gof and ROCR packages) [41].

This large registry-based study showed that the prevalence of statin use in secondary prevention increased strongly from 1999 to 2005, and more slightly from 2005 to 2013. Despite the strong increase in statin use in secondary prevention, there was rather a modest decline in the incidence of CVD, and this occurred mainly in older patients without statins prescribed. In 2013, less than 50% of the patients with a history of CVD received a statin. The absence of other secondary preventive medication was strongly associated with less statin use. Gender, age and comorbidity were associated with statin use to a lesser degree.

Analogous to previous studies, we found that the prevalence of statin use in secondary prevention has increased over the last few decades [9‐18, 20, 21, 23‐36, 42‐44]. The rise in statin use has been linked to guideline changes, promotions by the pharmaceutical industry, media reports, reimbursement conditions, entrance of generic medication and reductions of the price [10, 12, 15, 17, 18, 21, 24, 25, 28, 32, 33, 42, 43, 45].

This study computed AAPC of the prevalence of statin use in secondary prevention, which showed two trends: there was a sharp increase from 1999 to 2005 and a weaker, but also statistical significant increase from 2005 to 2013. This suggests that the steep increase in statin use in secondary prevention in the first time period was mainly linked to the growing literature in support of statins and to guideline changes. After 2005, when statins became less expensive for Belgian patients, the prevalence of statin use in secondary prevention rather seemed to reach a plateau. This finding is in line with previous studies [14, 15, 21, 24, 26, 27, 33, 36].

The current study showed that the incidence of recurrent CVD did not decrease much compared to the strong increase in the prevalence of statin use. Possible explanations are a higher prevalence of other cardiovascular risk factors and an inadequate control of other risk factors and perhaps the benefits of statins in most clinical trials do not translate to real world populations, which differ in ways that may adversely affect the risk-benefit balance (more comorbidities, polypharmacy, disability, etc). The current study also showed an increase in prevalence of diabetes, hypertension and hypercholesterolemia between 1999 and 2013. Unfortunately, insufficient data on the body mass index (BMI) and smoking status are available in the Intego registry. However, the Intego registry is representative for the Flemish population and the Belgian Health Interview Survey showed that the prevalence of overweight and smoking status in the age group 55+ years in Flanders did not change considerably from 1997 to 2013 (from 56.3 to 59% and from 17.6 to 16.9%, respectively) [46]. Moreover, the Euroaspire surveys described time trends in lifestyle, risk factor control and use of evidence-based medications in patients with coronary heart disease in Europe [14]. These surveys concluded that there was an increase of obesity and diabetes and that the proportion of smokers and the level of physical activity had remained stable from 1999 to 2013. Other studies showed similar results of cardiovascular risk factor control in secondary prevention [22, 28, 47‐49].

The minor decrease of the incidence of recurrent CVD was only observed in patients without statins prescribed. The incidence in this group was higher in 1999 and evolved towards the lower incidence of recurrent CVD in patients with statins prescribed in the 15 years thereafter, to be around 25 per 1000 patient years. However, since this study was only an observational study, both groups cannot be considered identical. Therefore, no conclusions could be drawn about the possible effect of statins on the incidence of recurrent CVD. This study was only designed to observe trends and generate hypotheses. Furthermore, no data were available on other factors that might have led to improved CVD outcomes in the same time period, like a reduction in trans fat consumption or attention to the treatment of sleep apnea.

Despite the strong increase in statin use in secondary prevention, the current study confirmed, in line with the results of previous studies, that less than 50% of patients in secondary prevention received a statin in 2013 [9‐18, 20, 21, 23‐36, 42‐44].

The absence of other secondary preventive medications (aspirin, RAS and non-RAS antihypertensive agents) was strongly associated with less statin use in secondary prevention. Possible explanations of these findings could be doctor-related factors such as poor knowledge and application of the guidelines, preference against polypharmacy or the doctor’s reading of the evidence that may be at variance with the guidelines, and patient-related factors such as poor adherence to therapy. Previous studies also found an association between the use of statins, aspirin and antihypertensive agents [11, 23].

Women were less likely to receive statins in secondary prevention than men. This finding is also in line with previous studies [9, 12, 15, 19, 23, 29, 31‐34]. Some of these studies showed that women also received less intensive statin therapy [9, 27, 34]. A possible reason for this observation might be the fact that women have a higher risk of statin adverse effects [50]. Furthermore, although statin therapy was shown to be an effective intervention in the secondary prevention of cardiovascular events in women, there was no benefit on stroke and all-cause mortality in women [51].

In the age group 60–79 years, higher age was associated with more statin use, whereas it was associated with less statin use in the oldest age group. This pattern of age-related statin use was also seen in other studies [12, 15, 21, 32, 33]. Although the most recent guidelines do not recommend age limitations for the use of statins in secondary prevention, current evidence does not support a favorable risk-benefit balance for statins in older persons [32]. Statins have shown to reduce the risk of coronary disease in older persons (70–82 years), but failed to reduce the risk for all-cause mortality and showed a statistically significant 25% increase in incident cancer [52]. Moreover, in the oldest old no trials have been performed and higher cholesterol concentrations have even been associated with longevity in this age group [53]. Furthermore, older age has also been linked to a greater risk of statin adverse effects [50].

The current study showed several comorbidities or cardiovascular risk factors were associated to statin use to a lesser degree. First, hypercholesterolemia was associated with more statin use in secondary prevention. This suggests that physicians prescribe statins as a function of cholesterol levels, despite the fact that statins are recommended in secondary prevention independently of cholesterol level [19, 20, 25, 32]. Second, diabetes predicted greater statin use in the age group 60–79 years. A possible explanation for this is the entrance of a diabetes care program in Belgium, with improved follow-up of these patients and increased awareness of recurrent CVD among patients with diabetes. Furthermore, diabetes has been classified as a coronary heart disease risk equivalent and as a stroke risk equivalent [54]. On the other hand, statin therapy has also been shown to be associated with a slightly increased risk of development of diabetes [55]. Moreover, statins have shown to raise glucose preferentially in patients at risk for diabetes, and to have increased occurrence of adverse effects in patients with diabetes [56]. Third, hypertension was associated with less statin use, possibly because patients and doctors are reluctant to polypharmacy. It has also been shown that geriatric patients of physicians who on average prescribed more medications like cholesterol lowering dugs, had an increased risk of mortality [57]. On the other hand, Xi et al. found that patients with hypertension after stroke are more likely to receive a statin [25]. Furthermore, the ALLHAT-LLT trial showed that pravastatin did not reduce either all-cause mortality or coronary heart disease in older persons with well-controlled hypertension [58]. Fourth, the observed associations between comorbidities like stroke, TIA, dementia and PAD and less statin use could possibly be explained by the higher risk of adverse effects [50, 59‐61], the limited effect of statins in people with these conditions [62, 63] and possibly the greater risk of drug errors in people with cognitive problems. Other interesting associations between less statin use and cancer and more statin use and paralysis should be confirmed by further research.

Based on the current findings future research should be organized. First, the effect of a statin treatment in secondary prevention could be estimated in this real world population by using propensity score matching, although this approach is not able to capture key issues of healthy user effects. Second, qualitative research could focus on the reason behind the finding that less than 50% of patients in secondary prevention receive a statin prescription. Although guidelines recommend statins in all age groups, clinicians might be aware of the failure of evidence to support net benefit in key patient groups, including elderly (age > 70 years), or might be reluctant to prescribe statins because of the risk for adverse effects. Furthermore, the impact of patient-related factors and patient preferences should be explored. Third, the current study suggests investigating what factors may be leading to declining CVD risk in older non-statin users.

The major strengths of the current study are the inclusion of a large real-world study population, representative of the general Flemish population, and the long follow-up period. This study was the first to compute AAPC of the prevalence of statin use in secondary prevention and to examine the prevalence of statin use parallel with the incidence of recurrent CVD in the same population. Furthermore, we were able to examine multivariable models to explain statin use in different age groups and have included a broad range of factors that may be associated with statin use.

However, this study also has limitations. First, this study was an observational study and statin users and non-users could not be considered identical. Therefore, no conclusions could be drawn about the causal effect of statins on the incidence of recurrent CVD. Second, no data were available on mortality. Third, only electronic GP prescriptions were taken into account. Manual prescriptions made during a house visit or prescriptions made by specialists were not included, which might underestimate the real prevalence of statin use. Fourth, we did not investigate the dosage and types of statin patients used and had insufficient data on actual cholesterol levels. Last, no information on smoking status, BMI, race, socio-economic status and time after diagnosis of the first CVD was available. Furthermore, secular trends that may alter CVD risk, like trans fat consumption or treatment of sleep apnea, were not examined.