Clinical decision-making for statin treatment in older patients with coronary artery disease (CAD) is under debate, particularly in community-dwelling frail patients at high risk of death. In this retrospective observational study on 2,597 community-dwelling patients aged ≥65 years with a previous hospitalization for CAD, we estimated mortality risk assessed with the Multidimensional Prognostic Index (MPI), based on the Standardized Multidimensional Assessment Schedule for Adults and Aged Persons (SVaMA), used to determine accessibility to homecare services/nursing home admission in 2005 to 2013 in the Padua Health District, Veneto, Italy. Participants were categorized as having mild (MPI-SVaMA-1), moderate (MPI-SVaMA-2), and high (MPI-SVaMA-3) baseline mortality risk, and propensity score–adjusted hazard ratios (HRs) of 3-year mortality rate were calculated according to statin treatment in these subgroups. Greater MPI-SVaMA scores were associated with lower rates of statin treatment and higher 3-year mortality rate (MPI-SVaMA-1 = 23.4%; MPI-SVaMA-2 = 39.1%; MPI-SVaMA-3 = 76.2%). After adjusting for propensity score quintiles, statin treatment was associated with lower 3-year mortality risk irrespective of MPI-SVaMA group (HRs [95% confidence intervals] 0.45 [0.37 to 0.55], 0.44 [0.36 to 0.53], and 0.28 [0.21 to 0.39] in MPI-SVaMA-1, -2, and -3 groups, respectively [interaction test p = 0.202]). Subgroup analyses showed that statin treatment was also beneficial irrespective of age (HRs [95% confidence intervals] 0.38 [0.27 to 0.53], 0.45 [0.38 to 0.54], and 0.44 [0.37 to 0.54] in 65 to 74, 75 to 84, and ≥85 year age groups, respectively [interaction test p = 0.597]). In conclusion, in community-dwelling frail older patients with CAD, statin treatment was significantly associated with reduced 3-year mortality rate irrespective of age and multidimensional impairment, although the frailest patients were less likely to be treated with statins.
There are limited data in relation to treatment with 3-hydroxy-3-methylglutaryl-CoA reductase inhibitors (statins) and mortality and other clinical outcomes in representative populations of community-dwelling frail older people with coronary artery disease (CAD). Frailty, a biologic syndrome reflecting a multidimensional state of decreased physiological reserve and increased vulnerability to stressors, has become a high-priority topic in cardiovascular medicine. Furthermore, recent guidelines have shown that the clinical decision-making on statin prescription in older patients only rarely is based on mortality risk stratification, resulting in many hospitalized or community-dwelling older patients with CAD not receiving statins. In older age, mortality risk stratification should be based on information on co-morbidity and functional status, and it is best performed using instruments based on multidimensional Comprehensive Geriatric Assessment (CGA), integrating information of several domains of health and function. Recently, a Multidimensional Prognostic Index (MPI) derived from a standardized CGA has been developed and validated for mortality risk assessment in several independent cohorts of hospitalized and community-dwelling older subjects with acute or chronic diseases. In a large sample of community-dwelling frail older patients with diabetes mellitus, statin treatment was associated with a reduced 3-year mortality rate, suggesting that a severely compromised health and functional status, or a very old age, did not affect the association between statin treatment and reduced mortality. The objective of the present study was to test the hypothesis that effectiveness of statin treatment in community-dwelling frail older patients with CAD may vary across strata of mortality risk in a 3-year follow-up period.
Methods
This was a retrospective observational study conducted according to the guidelines for Good Clinical Practice and the Strengthening the Reporting of Observational Studies in Epidemiology guidelines. All consecutive community-dwelling older subjects aged ≥65 years in whom a CGA-based multidimensional assessment using the Standardized Multidimensional Assessment Schedule for Adults and Aged Persons (SVaMA) was performed from January 1, 2005, to December 31, 2013, were screened for inclusion in the study. Patients were included in the analysis if they had been discharged from hospital with a main diagnosis of CAD according to International Classification of Diseases, Ninth Revision , 410 to 414 codes and subgroups or according to the main SVaMA diagnosis records K74 to K76 (Ischemic Heart Disease [IHD]) within 3 months from the SVaMA evaluation. The Institutional Review Board of the Social and Health-Care Local Unit (ULSS) 16, Padua, Italy, approved this study. Informed consent was given by participants who underwent SVaMA evaluation and/or by their proxies for clinical records to be used in clinical studies. The mean follow-up was 2.1 ± 2.2 years. The Registry Offices of cities where patients were residents at the time of the first evaluation were used to assess vital status during the follow-up, recording the dates of death from death certificates.
SVaMA is the instrument officially recommended since 2000 by the National Health Care System in the Veneto Regional Health System for multidimensional assessment by health professionals to establish accessibility of community-dwelling older persons to home care services or nursing home admission. The following SVaMA domains and variables were used for MPI calculation: (1) age, (2) gender, (3) main diagnosis, (4) nursing care needs (VIP) assessed by a validated numeric 11-item scale; (5) cognitive status (VCOG), assessed by the Short Portable Mental Status Questionnaire; (6) pressure sores risk (VPIA), assessed by the Exton-Smith Scale; (7) activities of daily living (VADL); (8) mobility (VMOB) assessed by the Barthel Index; and (9) social support (VSOC), assessed by a numeric 16-item scale that explores the presence of a support network during day and night. The following cut-off points were estimated for the normalized MPI-SVaMA 1-year mortality rate prediction: 0 to 0.33 (MPI-SVaMA-1 mild risk), 0.34 to 0.47 (MPI-SVaMA-2 moderate risk), 0.48-1.0 (MPI-SVaMA-3 severe risk). To calculate the MPI-SVaMA, software for Windows may be downloaded for free at the following address: http://www.mpiage.eu (English version). Further information on reliability, accuracy, calibration, and validation of the MPI based on the SVaMA can be found elsewhere. To extract the individual medication use, the whole study population was linked to the Pharmaceutical Prescription database of the Azienda ULSS 16, Padua. Statins and other drug prescriptions were determined according to the Anatomical Therapeutic Chemical codes. Statin prescription was determined by C10 Anatomical Therapeutic Chemical code. Subjects were considered statin users if they received statin prescriptions after the first registered evidence of the CAD diagnosis. In the present study, we included all statin users who achieved a treatment adherence coverage, that is, the ratio between treatment duration (in days) and individual follow-up duration (in days) of at least 90% for the first year and 80% and 70% when considering the outcome at 2 and 3 years of follow-up, respectively. We defined statin nonusers as the older subjects who never received statin prescriptions. We included prescriptions within 3 months after initial diagnosis of CAD. As a proxy of patients’ polypharmacy, we used the mean monthly past treatment rate defined as the total number of drug boxes taken before the enrollment divided by the total number of months between the first prescription and enrollment.
Frequencies (percentages) and mean restandard deviation (SD) were used to describe categorical and continuous baseline variables, respectively. Comparisons between men and women were performed using the Pearson chi-square test and Mann–Whitney U test, whereas linear trends across MPI-SVaMA risk subgroups were analyzed using analysis of variance models or the Mantel–Haenszel chi-square tests for continuous or categorical variables, respectively. Mortality rates were computed as the number of deaths per 100 person-years and compared using Poisson regression models. To control possible confounding effects on the association between statins treatment and mortality risk, the propensity score (PS) method was applied. PS logistic regression models were built to predict the probability of receiving statins according to all variables used for the calculation of MPI-SVaMA at treatment assignment: age, gender, VIP, VCOG, VPIA, VADL, VMOB, VSOC, the main diagnoses of fractures, cancer, dementia, stroke, hypokinetic syndrome and cardiovascular, respiratory, neurologic, or other diseases, and the past treatment rate of any drug (in tertiles). PS logistic models were selected stepwise, and model building was stopped when an adequate balance of covariates was achieved. Residual imbalances of covariates in PS quintiles were assessed at each step with a 2-way analysis of variance where each confounder was considered as an outcome and PS quintiles and treatment as factors. To verify that the data can support a comparison of treatment and control groups that are balanced on all covariates, the distribution of the estimated PSs for the treated and control groups should be checked for adequate overlap. This can be accomplished by creating overlapping histograms or by comparing quintiles of the estimated PSs for the treatment and control groups. If there is no overlap in the PS distribution across exposure groups, then no estimates of the treatment effect can be made. For this reason, subjects in treated and control groups with nonoverlapping PS distribution were excluded from the analysis. Separate PS logistic models were run for the overall sample and MPI-SVaMA subgroups. Multivariate and PS quintiles–adjusted Cox regression models were used to assess the effect of statins use on 3-year mortality rate, and results were reported as hazard ratios (HRs) and 95% confidence intervals (CIs). In addition, to check the robustness of our findings, a 5 to 1 greedy 1:1 PS-matching algorithm was performed. PS 1:1 matching identified a unique matched control for each patient treated with statins. Adequacy of covariate balance in the matched sample was eventually assessed with the McNemar or Wilcoxon signed rank test. For the overall sample and for specific MPI-SVaMA risk subgroups, adjusted HRs of statins use for 3-year mortality were reported along with numbers of events and subjects per group and mortality rates. Multivariate models included statin treatment, age, gender, main diagnoses, all domains of MPI-SVaMA, and the past treatment rates of any drug as covariates. As the PS-matched sample did not consist of independent observations, a marginal survival model with robust standard errors was used. p Values assessing the presence of a heterogeneous effect of statin treatment between MPI-SVaMA risk subgroups were also calculated and reported. Two-sided p values <0.05 were considered statistically significant. All analyses were performed using the SAS 9.1.3 statistical package (SAS Institute, Cary, North Carolina).
Results
Of a total population of 22,744 subjects aged ≥65 years who underwent a SVaMA evaluation over the study period, 3,172 (13.95%) were diagnosed with CAD. Of these, 334 and 241 subjects were excluded from analysis because of a time lag >3 months between CAD diagnosis and SVaMA evaluation or statin prescription, respectively. Thus, the final study population included 2,597 patients, 1,155 men (44.5%) and 1,442 women, with a mean age of 83.9 mo7.4 years. Men were younger than women (81.7 ea7.3 vs 85.7 856.8 years, p <0.001) and had higher MPI (0.37 ea0.1 vs 0.31 0.0.1, p <0.001), VIP (8.96 0.8.7 vs 7.01 .78.1, p <0.001), prevalence of cancer (25.8% vs 8.4%, p <0.001), and mortality rate over the 3-year follow-up (48.1% vs 34.5%, p <0.001). Women had a greater cognitive impairment than men (VCOG 5.52 ± 3.5 vs 4.80 ± 3.5, p <0.001) and a higher prevalence of dementia (26.4% vs 17.5%, p <0.001). Past drug treatment of any drug was similar in men and women, but the proportion of subjects starting statin treatment was greater in men than in women (46.5% vs 36.6%, p <0.001).
Table 1 lists the characteristics of older patients with CAD divided according to their MPI-SVaMA risk group: 785 (30.2%), 1,096 (42.2%), and 716 (27.6%) were at mild, moderate, and severe risk of mortality, respectively. Patients with greater MPI-SVaMA values were more likely to be men (p for trend <0.001) and older (p for trend <0.001) and had significantly greater VADL, VCOG, VIP, VMOB, VPIA, and VSOC scores (p for trend <0.001 for all domains). Three-year mortality rates were 23.4%, 39.1%, and 76.2% in the mild, moderate, and severe MPI-SVaMA risk subgroups, respectively (p for trend <0.001). Overall, 1,065 older patients with CAD (41.01% of the total study population) were treated with statins. Statin users were younger (p = 0.001) and included more men than statin nonusers (p <0.001; Table 2 ). Statin users had less impairment in VCOG (p <0.001), VPIA (p <0.001), VADL (p <0.001), and VMOB (p <0.001) scores, lower MPI-SVaMA values (p <0.001), and greater VIP values (p = 0.023) than statin nonusers. Moreover, statin users were more frequently in the MPI-SVaMA-1 group (35.6% vs 26.5%, p <0.001) and in the highest tertile of medication number than nonusers (3-tertile, 50.1% vs 22.1%, p <0.001).
