Statin Use Over 65 Years of Age and All-Cause Mortality

A 10-Year Follow-Up of 19 518 People

Sophia Eilat-Tsanani, MD; Elad Mor, MD; Yochai Schonmann, MD, MSc


J Am Geriatr Soc. 2019;67(10):2038-2044. 

In This Article


Setting and Data Sources

Israel operates a national health insurance scheme; residents may elect their healthcare provider and are assigned to a personal family physician.[25] In the northern region of Israel, more than 70% of the residents are members of Clalit Health Service (CHS). CHS operates an integrated electronic medical and administrative file, based on the International Classification of Diseases, 9th Revision. CHS also maintains a central register for chronic conditions. Diagnoses are based on reports of family physicians, community-based specialists, and hospital discharge letters. Diagnoses are cross-validated against medication possession records and laboratory data through an automated disease-specific process.[26–29] Members may fill their prescriptions at CHS-affiliated pharmacies at subsidized prices, facilitating the near-complete recording of medication possession.[30–33] Direct linkage to the Israel Ministry of Internal Affairs, using a unique national personal identification number, updates eligibility for exemption from copayments due to financial hardship or chronic disability and dates of death.

Study Population and Follow-up

We included all members of the Northern District of CHS who were aged 65 years or older on January 1, 2004, and were eligible for primary cardiovascular prevention (ie, those without a prior record of ischemic heart disease or stroke; Supplement S1). We used the data from the subsequent 2 years to ascertain baseline statin use and clinical characteristics. All participants entered the cohort on January 1, 2007, and they were followed up until the first of (1) date of death (any cause) or (2) end of the study period (December 31, 2016).

Baseline Characteristics

Baseline sociodemographic data included the date of birth, sex, ethnicity (Jewish or Arab), marital status, and eligibility for exemption from social security copayments. We included clinical information based on the date of inclusion in the CHS chronic disease register for the following conditions: smoking, obesity, hypertension, hyperlipidemia, diabetes, ischemic heart disease, cerebrovascular disease, peripheral vascular disease, carotid artery disease, valvular heart disease, chronic renal failure, and malignancy (Supplement S1). We considered diagnoses dates before cohort entry as baseline information and modeled subsequent entries (ie, after cohort entry) as time-updated variables.

Statin use

We retrieved the date, dose, and number of pills dispensed for all filled statin prescriptions between January 1, 2004, and December 31, 2016, using the World Health Organization Anatomical and Therapeutical Codes (Atorvastatin C10AA05, pravastatin C10AA03, rosuvastatin C10AA07, and simvastatin C10AA01). We considered those who had claimed any statin prescription during the baseline period (2004–2006) as "statin users." We expressed all statin prescriptions in terms of daily defined doses (DDDs) (1 DDD = atorvastatin 20 mg = pravastatin 30 mg = rosuvastatin 10 mg = simvastatin 30 mg).[34] We subsequently calculated the proportion of days covered (PDC) by statins for each participant by dividing the total amount of pills claimed by the total number of follow-up days (starting at cohort entry).[35] Because we used 1 DDD as the reference point, it was possible to achieve more than 100% adherence if participants were on higher intensity regimens (eg, a daily dose of 40 mg atorvastatin is equivalent to 2 DDDs). To capture changing patterns of statin use, we calculated participants' PDC for each calendar year separately (ie, the proportion of days covered by a statin between cohort entry and the last day of each calendar year, or the date of death if the participant had died during that specific year). Those who had claimed enough DDDs to cover 80% or more of their cumulative follow-up days at the end of each calendar year were considered adherent with statin therapy.[36,37] We modeled the cumulative adherence as a time-updated variable.

Study Outcomes

The primary outcome of the study was all-cause mortality. The secondary outcome was a "hard" ASCVD event, defined as the first occurrence of a fatal or nonfatal myocardial infarction or stroke.[6] The specific cause of death is not registered in the CHS database, but cardiovascular disease and malignancies are the two leading causes of death among older people in Israel, accounting for approximately one-half of all deaths, higher than any other specific cause by order of magnitude.[38] Therefore, to approximate cardiovascular-specific mortality we excluded from the analysis patients with any malignancy (censored on their date of malignancy diagnosis), an approach previously used to explore ASCVD events in Israel.[39,40]

Statistical Analysis

We presented the baseline characteristics by age group and by statin use at cohort entry. To assess the association between statin use and mortality and ASCVD, we fitted Cox regression models. We first considered potential covariates for inclusion in the regression models based on a literature review. We then applied a backward deletion change-in-estimate method for final model selection (simultaneously addressing multicollinearity and bias based on the calculated reduction of mean squared error of effect estimates).[41] First, we fitted a crude model, adjusting for age (using current age as a time scale) and sex. Subsequently, we fitted a minimally adjusted model, additionally adjusted for calendar period (2007–2011 vs 2012–2016), ethnicity, eligibility for social security waiver, and marital status. Finally, we fitted a fully adjusted model: also adjusted for time-updated smoking status, obesity, hypertension, hyperlipidemia, diabetes, cerebrovascular accidents, ischemic heart disease, carotid artery disease, peripheral vascular disease, congestive heart failure, cardiomyopathy, valvulopathy, chronic renal failure, and malignancy. To assess for interactions, we conducted stratified analyses by age and sex.

We conducted several sensitivity analyses to evaluate the robustness of some of our assumptions: (1) To explore potential bias by defining adherence as PDC of 80% or higher, we repeated the analysis, categorizing the PDC into narrower use bands (<20%, 20%-49%, 50%-79%, 80%-99%, 100%-119%, 120%-149%, and ≥150%); (2) we excluded from the analysis patients with diabetes (ie, diabetes as a "cardiovascular disease equivalent"); (3) we excluded from the analysis patients with renal failure, as defined in the CHD chronic disease register; (4) we defined adherence by the absolute number of pills dispensed (regardless of strength or dose) divided by follow-up days number of pills instead of DDDs; and (5) we repeated the analysis including only those who were statin users at baseline.

All P values reported are based on likelihood-ratio tests. The 95% confidence intervals were reported where appropriate. Statistical analysis was performed using Stata v.15.0 IC (StataCorp LP, College Station, TX). Ethical approval was obtained from the CHS institutional ethics review board (0058-16COM).