The authors have declared that no competing interests exist.
Conceived and designed the experiments: ADM HB HK WEH. Performed the experiments: ADM RQ JF KS BS BH DH. Analyzed the data: ADM. Contributed reagents/materials/analysis tools: ADM RQ JF KS BS DH. Wrote the paper: ADM RQ JF KS BS BH DH HK HB WEH.
Cardiovascular disease is a leading cause of death in older people, and the impact of being exposed or not exposed to preventive cardiovascular medicines is accordingly high. Underutilization of beneficial drugs is common, but prevalence estimates differ across settings, knowledge on predictors is limited, and clinical consequences are rarely investigated.
Using data from a prospective population-based cohort study, we assessed the prevalence, determinants, and outcomes of medication underuse based on cardiovascular criteria from Screening Tool To Alert to Right Treatment (START).
Medication underuse was present in 69.1% of 1454 included participants (mean age 71.1 ± 6.1 years) and was significantly associated with frailty (odds ratio: 2.11 [95% confidence interval: 1.24–3.63]), body mass index (1.03 [1.01–1.07] per kg/m2), and inversely with the number of prescribed drugs (0.84 [0.79–0.88] per drug). Using this information for adjustment in a follow-up evaluation (mean follow-up time 2.24 years) on cardiovascular and competing outcomes, we found no association of medication underuse with cardiovascular events (fatal and non-fatal) (hazard ratio: 1.00 [0.65–1.56]), but observed a significant association of medication underuse with competing deaths from non-cardiovascular causes (2.52 [1.01–6.30]).
Medication underuse was associated with frailty and adverse non-cardiovascular clinical outcomes. This may suggest that cardiovascular drugs were withheld because of serious co-morbidity or that concurrent illness can preclude benefit from cardiovascular prevention. In the latter case, adapted prescribing criteria should be developed and evaluated in those patients.
The predominant risks of a population with continuously increasing life expectancy include the simultaneous use of multiple medications [
In cardiovascular disorders, underuse is frequent and well documented for many drug classes [
The
The study population is a subsample of the ESTHER study, a large population-based cohort study conducted in Germany described in detail elsewhere [
In the home visit, 2714 participants provided medication records with at least one drug coded by the Anatomical Therapeutic Chemical (ATC) classification system. These participants were found eligible for the inclusion into our subsample. Those participants with a documented medical history of cardiovascular diseases according to adapted cardiovascular criteria from the START list from 2008 [
criterion | description |
---|---|
A3 | Antiplatelets with a documented history of atherosclerotic coronary, cerebral, or peripheral vascular disease |
A4 | Antihypertensive therapy where systolic blood pressure consistently exceeded 160 mmHg in repeated measurements during the home visit |
A5 | Statin therapy with a documented history of coronary, cerebral, or peripheral vascular disease |
A8 | Beta-blocker therapy in patients with chronic stable angina |
F3 | Antiplatelet therapy in diabetes mellitus with coexisting major cardiovascular risk factors |
F4 | Statin therapy in diabetes mellitus if coexisting major cardiovascular risk factors are present |
a a documented history of atherosclerotic coronary, cerebral, or peripheral vascular disease included previous myocardial infarction, stroke, coronary intervention (bypass surgery or balloon catheterization of the coronary arteries), pulmonary embolism, and deep vein thrombosis.
b hypertension, hypercholesterolemia, and smoking history
The participants’ exposure to indicated medications was determined during a brown bag medication review, in which all drugs were unequivocally identified and recorded electronically by the study physician during the home visit. Information on drug utilization was complemented by the medical history of prevalent diseases at the time of the home visit in order to determine underuse according to the criteria listed in
We defined relevant study outcomes as cardiovascular mortality, non-fatal myocardial infarction, stroke, or coronary interventions such as bypass surgery or balloon catheterization of the coronary arteries. Non-cardiovascular mortality, i.e. death due to non-cardiovascular causes, was the competing event in competing risk analyses. We considered major changes in health status as concurrent events in a time-dependent covariate because they might provoke a reconsideration of the treatment or, more generally, might influence a participant’s covariate status. As such, we chose incident diagnoses of coronary heart disease, heart failure, renal failure, and pulmonary embolism.
Concerning covariates, co-morbidity was defined by the Cumulative Illness Rating Scale for Geriatrics (CIRS-G) as reported by the responsible GP. Information on income classes, presence of a partner or cohabitant, number of social contact persons, educational level (basic: ≤9, middle: 10–11, higher: ≥12 years), and smoking status (current, past, or never smoker) was extracted from the participant’s questionnaire. Physical examination during the geriatric assessment included two blood pressure measurements from both arms. The body mass index (BMI) was calculated based on weight and height values measured during the home visit. Calibrated devices were used to measure blood pressure, height, and weight. Validated assessment tools provided information on cognitive status (MMSE) and performance in activities of daily living (Barthel Index). We applied a modified phenotypic definition of frailty by Fried and co-workers [
Vital status was ascertained via local population registries and information on the cause of death of fatal events were provided by public health departments providing the underlying cause of death according to the International Classification of Diseases, 10th revision [ICD-10] with cardiovascular deaths being defined by ICD-10 codes I00–I99.
The study was approved by the Ethics Committees of the Medical Faculty of the University of Heidelberg (Study-ID: 058/2000) and of the Medical Board of the State of Saarland (Study-ID: 67/00) and was conducted in accordance with the declaration of Helsinki. Written informed consent was obtained from each participant prior to the study.
Standard descriptive methods were used to describe demographic characteristics of the study sample at baseline. Between-group differences were assessed using the Chi-square test for categorical variables and either the Student’s t-test or the Wilcoxon rank-sum test for ordinal or continuous variables, depending on their distribution.
Potential factors associated with medication underuse according to START criteria were determined in a logistic regression analysis. We applied variable selection by blocks: in the first block, variables with proven association (cognitive status and number of drugs [
We conducted time-to-event analyses to study the potential impact of medication underuse on relevant cardiovascular events and competing deaths due to non-cardiovascular causes. Non-parametric methods using the log-rank test or the Gray test for cumulative incidence functions were applied in preliminary descriptive analyses. In the main analysis, we modeled cause-specific hazards while accounting for major concurrent health events as a time-dependent covariate. Besides mandatory covariates for adjustment (age, sex, co-morbidity, and presence of cardiovascular risk factors), we included variables with a significant association with the prescription status determined by the logistic regression model. The assessment of the proportional hazards assumption was checked and considered in the model building by means of Schoenfeld residuals. We further investigated sensitivity of our results by considering all-cause mortality as a part of our composite endpoint of relevant outcomes or applying multiple imputation of missing values.
All tests were two-tailed, 95% confidence intervals (CI) were calculated, and
From the available 2714 medication records, 1454 (53.6%) participants fulfilled the inclusion criteria. The mean age (± standard deviation, SD) of included study participants was 71.1 ± 6.1 years (range 58–84) and 672 (46.2%) were female. Underuse defined as at least one missing medication was present in 1005 participants (69.1%). The demographic, clinical, medication-related, and socio-economic characteristics of the participants stratified for underuse or appropriate use revealed group differences at baseline for age, frailty, independence of daily life, and the number of drugs (
Variables | Underuse | Appropriate use | All | ||||
---|---|---|---|---|---|---|---|
N | value | N | value | N | value | Subgroupcomparison | |
female | 481 | (47.9) | 191 | (42.5) | 672 | (46.2) | |
male | 524 | (52.1) | 258 | (57.5) | 782 | (53.8) | |
total | 1005 | 449 | 1454 | ||||
<65 | 199 | (19.8) | 62 | (13.8) | 261 | (18.0) | |
65–74 | 565 | (56.2) | 273 | (60.8) | 838 | (57.6) | |
>75 | 241 | (24.0) | 114 | (25.4) | 355 | (24.4) | |
Mean ± SD | 70.8 ± 6.2 | 71.7 ± 5.8 | 71.1 ± 6.1 | ||||
Mean (Med) | 8.5 (7) | 8.8 (8) | 8.5 (7) | ||||
Mean ± SD | 29.9 ± 5.2 | 29.4 ± 4.8 | 29.7 ± 5.1 | ||||
Mean (Med) | 27.9 (28) | 28.1 (29) | 27.9 (29) | ||||
non-frail | 250 | (24.9) | 131 | (29.2) | 381 | (26.2) | |
pre-frail | 598 | (59.5) | 267 | (59.5) | 865 | (59.5) | |
frail | 150 | (14.9) | 47 | (10.5) | 197 | (13.6) | |
Mean (Med) | 98.0 (100) | 98.8 (100) | 98.2 (100) | ||||
current | 73 | (7.3) | 38 | (8.5) | 111 | (7.6) | |
former | 433 | (43.1) | 199 | (44.3) | 632 | (43.5) | |
non-smoker | 483 | (48.1) | 201 | (44.8) | 684 | (47.1) | |
<5 | 407 | (40.5) | 100 | (22.3) | 507 | (34.9) | |
5–9 | 543 | (54.0) | 308 | (68.6) | 851 | (58.5) | |
>10 | 55 | (5.5) | 41 | (9.1) | 96 | (6.6) | |
Mean (Med) | 5.5 (5) | 6.6 (6) | 5.8 (6) | ||||
N (%) | 55 | (5.5) | 37 | (8.2) | 92 | (6.3) | |
Mean (Med) | 4.7 (5) | 4.8 (5) | 4.8 (5) | ||||
Mean (Med) | 4.4 (4) | 4.2 (3.5) | 4.3 (4) | ||||
basic | 695 | (69.2) | 297 | (66.1) | 992 | (68.2) | |
middle | 155 | (15.4) | 76 | (16.9) | 231 | (15.9) | |
higher | 134 | (13.3) | 73 | (16.0) | 206 | (14.2) |
Med: Median; SD: standard deviation
a Ordinal income classes per month were calculated as follows 1: less than 500 Euro; 2: 500 to 750 Euro; 3: 750 to 1000 Euro; 4: 1000 to 1500 Euro; 5: 1500 to 2000 Euro; 6: 2000 to 3000 Euro; 7: 3000 to 5000 Euro; 8: more than 5000 Euro
b Classification: “basic” <9 years; “middle” 10–11 years; “higher” >12 years
In our subsample, more than two thirds of the participants had a documented history of atherosclerotic coronary, cerebral, or peripheral vascular disease. Among the 981 participants with a documented history of atherosclerotic disease, 50.9% did not use antiplatelet agents and 51.1% did not use a statin. Likewise, antiplatelet agents were missing in 50.4% and statins were withheld in 52.7% of the 546 participants with diabetes and at least one additional major cardiovascular risk factor. No beta-blocker was used by 31.3% of the 467 participants with chronic stable angina, and no antihypertensive drugs were used by 20.4% of the 270 participants with a systolic blood pressure above 160 mmHg.
To assess predictors of medication underuse in a multivariate way, we considered all variables listed in
In the final model, statistical significance was observed for frailty, BMI, and the number of drugs (
Selected model variables and their association with medication underuse in a multivariate logistic regression model (*** < 0.001; ** < 0.01; and * < 0.05).
1005 participants with medication underuse (77 lost to follow-up, 8.3%) and 449 appropriately treated participants (36 lost to follow-up, 8.7%) contributed a mean follow-up time of 2.24 years with a total number of 140 cardiovascular events. Neither a Kaplan-Meier plot (
(A) Kaplan-Meier plot of relevant cardiovascular events for appropriate use and medication underuse (
In the multivariate regression model, medication underuse was not associated with cardiovascular outcomes (HR = 1.00, CI [0.65, 1.56],
Full Analysis Set | Subgroup ≥ 65 years | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Parameter |
Relevant events | Competing event | Relevant events | Competing event | ||||||||
HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | |||||
Underuse | 1.00 | 0.65, 1.56 | 0.987 | 2.52 | 1.01, 6.30 | 0.047 | 1.17 | 0.72, 1.91 | 0.516 | 3.63 | 1.23, 10.7 | 0.019 |
Health event | 1.49 | 0.64, 3.43 | 0.354 | 1.80 | 0.78, 4.18 | 0.170 | ||||||
BMI [kg/m2] | 0.96 | 0.91, 1.00 | 0.057 | 0.97 | 0.90, 1.04 | 0.372 | 0.95 | 0.90, 1.00 | 0.070 | 0.99 | 0.91, 1.07 | 0.734 |
Co-morbidity | 1.05 | 1.02, 1.08 | 0.004 | 1.08 | 1.03, 1.13 | 0.002 | 1.06 | 1.02, 1.09 | 0.002 | 1.08 | 1.03, 1.13 | 0.001 |
CVD risk factor |
2.79 | 0.68, 11.5 | 0.156 | 2.60 | 0.34, 19.9 | 0.358 | 2.54 | 0.61, 10.5 | 0.198 | 2.60 | 0.34, 20.1 | 0.360 |
Frailty | 1.04 | 0.72, 1.50 | 0.838 | 1.75 | 0.96, 3.20 | 0.068 | 0.93 | 0.62, 1.38 | 0.708 | 1.69 | 0.90, 3.16 | 0.101 |
(BMI: body mass index; CI: confidence interval; CVD: cardiovascular disease; HR: hazard ratio)
a Warranting the assumption of proportional hazards, the model was additionally stratified for sex, age groups, and categories of drug numbers as indicated by Schoenfeld residuals.
b operationalized as a dichotomous variable indicating the presence of any cardiovascular risk factor (hypertension, hypercholesterolemia, and smoking history
c no estimates are reported due to shortage of events leading to imprecise estimates with confidence intervals ranging to infinity
This subsample of the ESTHER study with a mean age of 71 years showed that medication underuse was common (69.1%), even though we enrolled only participants taking at least one drug. The prevalence is high compared to other studies [
With respect to factors potentially influencing medication underuse, we confirmed earlier reported predictors of underuse such as sex [
We prospectively assessed the association of underutilization of preventive medications with cardiovascular outcomes. Only one study [
The subgroup analysis of participants aged ≥65 years indicated an age-dependent effect of medication underuse. This presumed finding was sustained by the association detected for age as a continuous variable in the Cox regression model, clearly emphasizing that medication underuse becomes strikingly more important with increasing age: The hazard ratio of the undertreated population compared to the appropriately treated population increased by 7% with each year of life. This finding corroborates the age requirement ≥65 years as stated in the START criteria [
In our population, medication underuse was unrelated to cardiovascular events of patients with cardiovascular disease, but was associated with a higher death rate from non-cardiovascular causes. A closer look at the actual death codes revealed that non-cardiovascular deaths were often reported as cancer. Although this does not imply that these patients ultimately died from cancer as opposed to cardiovascular causes, it might suggest a refusal to apply preventive cardiovascular medications in view of the participants’ co-morbidity. Frailty as an indicator of worse survival [
Especially in everyday practice individual characteristics of a patient have to be considered, of which many would have led to the exclusion from the selected population of a randomized trial. Because (frail) older patients with a limited life expectancy are hardly ever included in those trials, our unexpected findings are especially valuable not only in the exploration of the preventive effectiveness of recommended medications, but also in the recognition of required further research with frail older patients who were often not adequately treated. The results raise the question of effectiveness in special populations such as frail older people.
The following potential limitations of our study design are worth to be considered. While the use of brown bag medication review adequately addressed the problem of underreporting, we had to extrapolate the stability of medication use over time and thus persistence and long-term exposure. Besides, further baseline variables may have changed over time. We took this issue into account by including major incident diagnoses as a time-dependent covariate reflecting the reason for potential changes in baseline variables such as the medication status. Finally, a larger sample size and a longer follow-up time would have been desirable. Considering our effect size of a literal null effect, no power recalculation was conducted.
In conclusion, the thorough evaluation of actual medications and prevalent diagnoses in an aging ambulatory population with cardiovascular disease revealed that a sizeable fraction of this population met the criteria for underuse of cardiovascular drugs. Patients were less likely to use such medications if they were overweight, frail, and if they were prescribed less medicines. Medication underuse was not associated with increased risk for cardiovascular events but was associated with an increased risk for non-cardiovascular mortality. This may suggest that, at times, cardiovascular drugs are withheld in patients because of concurrent serious co-morbidity that potentially precludes benefit for pharmacological cardiovascular prevention. Future research is necessary to explore therapeutic effectiveness of preventive cardiovascular drugs in special populations such as frail older people.