Beta‐blocker use and fall risk in older individuals: Original results from two studies with meta‐analysis

Abstract

Aims

To investigate the association between use of β‐blockers and β‐blocker characteristics – selectivity, lipid solubility, intrinsic sympathetic activity (ISA) and CYP2D6 enzyme metabolism – and fall risk.

Methods

Data from two prospective studies were used, including community‐dwelling individuals, n = 7662 (the Rotterdam Study) and 2407 (B‐PROOF), all aged ≥55 years. Fall incidents were recorded prospectively. Time‐varying β‐blocker use was determined using pharmacy dispensing records. Cox proportional hazard models adjusted for age and sex were applied to determine the association between β‐blocker use, their characteristics – selectivity, lipid solubility, ISA and CYP2D6 enzyme metabolism – and fall risk. The results of the studies were combined using meta‐analyses.

Results

In total 2917 participants encountered a fall during a total follow‐up time of 89 529 years. Meta‐analysis indicated no association between use of any β‐blocker, compared to nonuse, and fall risk, hazard ratio (HR) = 0.97 [95% confidence interval (CI) 0.88–1.06]. Use of a selective β‐blocker was also not associated with fall risk, HR = 0.92 (95%CI 0.83–1.01). Use of a nonselective β‐blocker was associated with an increased fall risk, HR = 1.22 (95%CI 1.01–1.48). Other β‐blocker characteristics including lipid solubility and CYP2D6 enzyme metabolism were not associated with fall risk.

Conclusion

Our study suggests that use of a nonselective β‐blocker, contrary to selective β‐blockers, is associated with an increased fall risk in an older population. In clinical practice, β‐blockers have been shown effective for a variety of cardiovascular indications. However, fall risk should be considered when prescribing a β‐blocker in this age group, and the pros and cons for β‐blocker classes should be taken into consideration.

What is Already Known about this Subject

• Beta‐blocker use has been associated with fall risk, although literature is contradictory.

• Pharmacological and adverse effects may vary between β‐blocker characteristics.

• Therefore, the association between β‐blocker characteristics – adrenergic receptor selectivity, lipid solubility, intrinsic sympathetic activity (ISA) and CYP2D6 enzyme metabolism – and fall risk should be evaluated.

What this Study Adds

• Use of a nonselective β‐blocker, in contrast to selective β‐blockers, is associated with an increased fall risk in an older population.

• Lipid solubility and CYP2D6 enzyme metabolism were not associated with fall risk.

• The number of participants using a β‐blocker with ISA was limited and therefore an association with fall risk could not be examined.

Introduction

In the aging population, fall incidents form a growing healthcare problem 1. Of those older than 65 years, one in three encounters at least one fall annually 2. Falls lead to significant morbidity and even mortality. Moreover, falls are associated with reduced quality of life and increased health care costs 3, 4, 5. One of the risk factors for falls is the use of certain medication 6, 7, including β‐blockers, although literature is contradictory 7, 8, 9, 10. Beta‐blocker use is thought to result in fall risk by inducing bradycardia, reducing the cardiac output, inducing hypotension and dizziness 11. Pharmacological effects and occurrence of adverse effects may vary between different β‐blocking agents. Differences of β‐blocking agents relate for example to their selectivity for adrenergic receptors, lipid solubility, intrinsic sympathetic activity (ISA) and their elimination route 11, 12.

Previously, we observed an increased fall risk with the use of nonselective β‐blocking agents 8. In addition, the more lipophilic β‐blockers may be associated with central nerve system side effects, such as dizziness and light‐headedness 11, 13. Furthermore, β‐blocking agents with ISA might be less susceptible to cause bradycardia 11. Regarding the elimination route, some β‐blockers are eliminated through liver metabolism (e.g., metoprolol and propranolol), whereas others are predominantly eliminated by renal excretion (e.g., atenolol) 11, 12. For those subjected to liver metabolism, the Cytochrome P450 (CYP) 2D6 enzyme plays an important role 14, 15. The CYP2D6 gene displays multiple genetic variations, of which the *4 variant allele is suggested to be of main importance for Caucasians. The *4 variant results in a nonfunctional protein 16, and in Caucasians it is responsible for the majority of poor metabolizer phenotypes 14, 15. Previous research indicated that metoprolol users with a poor metabolizers phenotype, according to their CYP2D6*4 genotypes, had a lower blood pressure and were at increased risk for bradycardia 17, 18. Overall, varying pharmacological effects of β‐blocking agents or individual differences on clearance may underlie the contradictory literature results.

Our objective was to investigate the association between use of β‐blocker and β‐blocker characteristics – selectivity, lipid solubility, ISA and CYP2D6 enzyme metabolism – and fall risk. We hypothesized that use of nonselective agents, lipid soluble agents and those without ISA are associated with an increased fall risk. In addition, we hypothesized that users of β‐blockers metabolized by CYP2D6 who carry a CYP2D6*4 variant are also at increased risk for fall incidents. These research questions were investigated in two independent studies involving community‐dwelling older individuals.

Methods

Study population and setting

Data were used from the Rotterdam Study and B‐PROOF (B‐vitamins in the PRevention Of Osteoporotic Fractures). The Rotterdam Study is an ongoing population‐based cohort, executed within a suburb of Rotterdam. Its design, objectives and methods have been described in detail 19, 20. Briefly, the study was initiated in 1989 and 7983 participants aged ≥55 years were included. Subsequently, participants were interviewed and underwent an extensive set of examinations that were repeated during the follow‐up visits every 4–5 years. For the current study, participants with pharmacy dispensing data and validated fall data were included, covering a study period from 1 May 1991 until 31 December 2010. The Rotterdam Study has been approved by the medical ethics committee according to the ‘Wet Bevolkingsonderzoek ERGO’ (Population Study Act: Rotterdam Study), executed by the Ministry of Health, Welfare and Sports of the Netherlands. All study participants gave written informed consent to participate in the study and obtain information from their treating physicians 19, 20.

B‐PROOF has also been described in more detail 21. In short, it is a multicentre, randomized, placebo‐controlled, double‐blind trial investigating the efficacy of vitamin B₁₂ and folic acid supplementation on the prevention of fractures in persons aged ≥65 years. In total, 2919 participants were included and followed for 2–3 years, covering a study period from 2008 until 2013. Inclusion criteria were homocysteine levels of 12–50 μmol l^–1, serum creatinine ≤150 μmol l^–1, and no reported malignancies in the past 5 years. For the current study participants with pharmacy dispensing data were included. The Medical Ethics Committee of Wageningen University approved the study protocol, and the Medical Ethics committees of Erasmus Medical Centre and VU University Medical Center gave approval for local feasibility. Before entering the study, all participants gave written informed consent 21.

Previous B‐PROOF results indicated that the intervention had no effect on the time to first or second fall, or the number of falls encountered during the study 22. For the current study we therefore used a cohort study design.

Fall incidents

In the Rotterdam Study, serious falls were defined as ‘a fall leading to a hospital admission or leading to a fracture’. Data were obtained from a computerized reporting system of the general practitioners within the Rotterdam Study. Participant data were also linked to the Dutch National Morbidity Registration, which contains information of all hospital admissions. Serious fall data were coded by two members of the research team and were completed until 2010. The first serious fall date was defined as the index date. A participant was followed from the baseline date (date of study enrolment) until the first serious fall (index date), death or the end of the study period, whichever came first.

In B‐PROOF, a fall incident was defined as ‘an unintentional change in position resulting in coming to a rest at a lower level or on the ground’ 23. Participants reported fall incidents prospectively on a fall calendar on a weekly basis. The calendar was returned to the research team every 3 months. Participants with incomplete or unclear calendars were contacted by telephone. Participants were followed until their first fall incident; the Thursday in that particular week was defined as the index date. Participants were followed from baseline until the index date, their drop‐out date or the date of their last calendar, date of death, or the end of the study, whichever came first.

Beta‐blocker use

Beta‐blocker use was defined according to the Anatomical Therapeutic Chemical (ATC) code 24, C07. Selective β‐blockers were defined with the ATC codes: C07AB, C07BB, C07CB, C07DB, C07EB and C07FB. Nonselective β‐blockers were defined as C07AA, C07BA, C07CA, C07DA, C07EA, C07FA and C07AG. Table 1 lists the characteristics of individual β‐blocking agent, according to lipid solubility, ISA, and CYP2D6 enzyme metabolism 11, 12.

Table 1.

The characteristics of individual β‐blocking agent, according to lipid solubility, intrinsic sympathetic activity (ISA) and CYP2D6 enzyme metabolism

Beta‐blocker agent	Selectivity	Lipophilicity	ISA	CYP2D6 metabolism
Acebutolol	x	x	x
Alprenolol		xx	x	x
Atenolol	x
Betaxolol	x	x
Bevantolol	x	x
Bisoprolol	x	x
Carteolol			x
Carvedilol		xx		x
Celiprolol	x		x
Labetalol		x
Metoprolol	x	x		x
Nebivolol	x	x		x
Oxprenolol		x	x
Penbutolol		xx	x
Pindolol		x	x
Propranolol		xx		x
Sotalol
Timolol		x		x

‘x’ indicates the presence of a characteristic, and ‘xx’ indicates the highly lipophilic agents

Five exposure definitions were used in the analyses: i) β‐blocker use overall; ii) selective‐ and nonselective β‐blocker use; iii) lipophilic and nonlipophilic β‐blocker use; iv) β‐blockers with and without ISA; and v) use of β‐blockers with and without CYP2D6 enzyme metabolism.

In both studies, β‐blocker use was based on pharmacy dispensing records. Thereby, we were able to use β‐blocker exposure as a time‐varying determinant. In the Rotterdam Study, these records from the regional pharmacies were available from 01 January 1991 onwards. More than 95% of the participants fill their drug prescriptions at one of these pharmacies.

In B‐PROOF, pharmacy dispensing records were obtained from the Dutch Foundation for Pharmaceutical Statistics, as previously described 8. Data were available throughout the study period of a participant.

The dispensing records contain information regarding date of dispensing, total amount of drug units per dispensing, prescribed daily number of units, product name of the drugs and corresponding ATC code. Current medication use was defined as use at the time of the fall (on the index date). Past use was defined as use prior to, but no longer on, the index date. To investigate a dose–response relation, the average prescribed daily dose was expressed in standardized defined daily doses (DDDs). In the Rotterdam Study, we ensured that all participants had pharmacy dispensing records available for at least 4 months prior to their study start, to avoid potential misclassification of exposure.

Covariables

Basis characteristics including age, sex, ethnicity, use of a walking aid, history of falls, smoking habits, alcohol consumption and diabetes were ascertained using a questionnaire 19, 21. During study visits, various characteristics were measured including weight, height, blood pressure, depressive symptoms and cognitive performance. Additionally, serum creatinine 19, 21, and the use of concomitant medication were assessed. As a measure of physical function, lower limb disability scores were determined in the Rotterdam Study 25, 26, in B‐PROOF physical performance scores and hand‐grip strength were assessed 27, 28. In the Rotterdam Study, also orthostatic hypotension measures and dizziness were available 19.

In the Rotterdam Study, alcohol consumption was based on food frequency data and reported in g day^–1 29. In B‐PROOF, alcohol consumption was categorized into light, moderate and excessive 30. Diabetes was based on self‐report 19, 21. Weight and height were measured and were used to calculate body mass index (kg m^–2). Hypertension was defined as systolic blood pressure > 140 mmHg and/or diastolic blood pressure > 90 mmHg 31. In the Rotterdam Study, depressive symptoms were assessed using the Center for Epidemiological Studies Depression Scale (score range 0–60) 32 or in a subsample the Hospital Anxiety and Depression Scale (score range 0–21) 33. In B‐PROOF, the 15‐item version of the geriatric depression scale was used 34. Clinically relevant depressive symptoms were based on Center for Epidemiological Studies Depression Scale scores ≥16 35, 36, Hospital Anxiety and Depression Scale scores ≥9 37, or geriatric depression scale scores ≥5 38. Cognitive performance was assessed by the mini‐mental state examination 39. Serum creatinine levels were used to calculate an age‐adjusted estimate of the glomerular filtration rate according to the chronic kidney disease epidemiology collaboration formula 40. Concomitant medication use was assessed with pharmacy dispensing records, and those considered as potential confounders were; antihypertensive medication C02, diuretics C03, calcium antagonist C08, renin‐angiotensin agents C09, benzodiazepines N05BA or N05CD, and antidepressants N06A. Lower‐limb disability scores were assessed using a modified version of the Stanford Health Assessment Questionnaire 25, 26. The score was based on answers to questions regarding rising, walking, bending, and getting in and out of a car. Disability was defined as a score of 3 or higher 26. A physical performance score was calculated from the results of three physical function tests: walking test, chair stand test, and the tandem stand test 28 Physical performance score ranged from 0–12 (low physical performance–high physical performance) 27, 41. Maximum handgrip strength (kg) was defined as the highest results of two maximum trials per hand using a dynamometer (Takei TKK 5401; Takei Scientific Instrument Co. Ltd., Tokyo, Japan). Orthostatic hypotension was defined as: a decrease of ≥20 mmHg in systolic and/or a decrease of ≥10 mmHg in diastolic blood pressure 42. Dizziness symptoms were assessed using a questionnaire 19.

For the Rotterdam Study, baseline fall history, dizziness and serum creatinine levels were used, and depressive symptoms were available from the second visit onwards. Data of the other covariables were available for all follow‐up visits, but to minimize the number missing, the values from the preceding visit were used.

Genotyping

CYP2D6*4 (rs3892097) allele variants were determined based on the Illumina 550 (+duo; the Rotterdam Study), and the Illumina‐Omni express array (B‐PROOF), and imputation to 1000 Genomes Project (PhaseIv3, March 2012) reference set 43. The imputation quality was 0.99. In B‐PROOF imputations were only done for Caucasians. For both studies, the reference group was defined as those with homozygous major allele carriership (i.e. the absence of a CYP2D6*4 allele).

Statistical analyses

Baseline characteristics were determined for fallers and nonfallers. Differences between groups were tested using a t test, a Chi‐square test or a Mann–Whitney U test. Deviation from Hardy–Weinberg equilibrium was tested using a Chi‐square test for allele frequencies.

Cox proportional hazards models were used to calculate fall hazard ratios (HR) for users compared to nonusers 44. The model compares the prevalence of exposure to β‐blockers in the incident fall cases on the index date with the exposure prevalence in all other participants in the cohort on the same date of follow‐up. In this way, cases are censored but noncases can serve as a reference on multiple occasions until the end of the study period. This method for cohort analysis with a Cox proportional hazards analysis with drug use as a time‐varying determinant is valid and has been described earlier 44. This analysis was done for the five exposure categories (i.e., β‐blocker use overall, selectivity, lipophilicity, ISA and CYP2D6 enzyme metabolism) separately. Per exposure category we stratified on β‐blocker characteristic (e.g., nonuse vs. selective‐ and nonselective β‐blocker use). Nonuse was defined as no current β‐blocker use. The models were adjusted for age and sex (model 1). Covariables were included in the models if they changed the hazard ratio of the association between β‐blocker use and falls by > 10% (model 2). For the first exposure, a dose–response relation was investigated. Dose categories were made according to median number of prescribed DDDs. In addition, the analysis of the fifth exposure was stratified on CYP2D6 genotypes. Fall risk in current users was compared to nonusers, within those carrying no variant CYP2D6*4 alleles. Likewise, within those carrying at least one variant CYP2D6*4 allele, fall risk in current users was compared to nonusers.

The results of the two studies were combined using meta‐analysis. The effect estimates – β's (log HR) – and their standard errors were used to calculate the overall effect, and to investigate the heterogeneity between the studies. Meta‐analyses were done using the R package ‘rmeta’ applying a random effect model, R version 3.0.3. All other statistical analyses were done using the statistical software package SPSS version 21.0 (IBM, Armonk, NY, USA) and P‐values <0.05 were considered statistically significant.

Finally, sensitivity analyses were applied, in which we categorized the selectivity β‐blocker group into: no β‐blocker use (reference); selective β‐blocker use; nonselective; and nonselective and also high lipophilic β‐blocker use. The lipophilic β‐blocker group was categorized into: no β‐blocker use (reference); nonlipophilic; medium lipophilic; and highly lipophilic. In addition, a dose–response relation was investigated for nonselective β‐blockers. Furthermore, an association with falls for past and current use was investigated for selective and nonselective β‐blockers.

Results

Study population

The total Rotterdam Study population included 7983 participants. Of those, 7662 had both medication and fall data, with 6170 having also genetic data. B‐PROOF included 2919 participants, of whom 2407 had medication and fall data, and 2135 also had genetic data (flow‐chart Figure S1). The Rotterdam population with medication, fall, and genetic data differed slightly from those without genetic data. Those without genetic data were slightly older and more likely to: be female; have a positive fall history; use a walking aid; tohave a lower mini‐mental state examination score; have depressive symptoms, hypertension, self‐reported diabetes or lower limb disabilities; and be of non‐Caucasian origin. In addition, there were fewer current smokers. In B‐PROOF, those without genetic data were more likely to have self‐reported diabetes.

In the Rotterdam Study the median follow‐up time was 11.4 year with an interquartile range of 5.1–17.9 years, and in B‐PROOF it was 1.8 years (0.5–2.0). Table 2 presents the baseline characteristics for both study populations, separated on occurrence of a fall during follow‐up.

Table 2.

Baseline characteristics of the Rotterdam Study and B‐PROOF grouped on the basis of fall incidents during follow‐up

	Rotterdam Study		B‐PROOF
	Fallers (n = 1770)	Nonfallers (n = 5892)	Fallers (n = 1147)	Nonfallers (n = 1260)
Age, years a	71.6 (9.3)	69.7 (9.5)d	74.4 (6.7)	73.7 (6.1)d
Female sex b	1360 (76.8)	3270 (55.5)d	617 (53.8)	564 (44.8)d
Caucasian ethnicity b	1640 (92.7)	5477 (93.0)	1080 (94.2)	1186 (94.1)
Body mass index a	26.3 (3.7)	26.3 (3.7)	26.9 (4.0)	27.3 (4.0)
History of falls (yes) b , e	392 (22.6)	907 (15.7)d	399 (44.2)	217 (21.4)d
Walking aid use (yes) b	215 (13.2)	591 (10.7)d	172 (15.1)	153 (12.2)d
MMSE score c	28 [26–29]	28 [26–29]	29 [27–29]	28 [27–29]d
Depressive symptoms (yes) b , f	153 (13.6)	339 (9.0)d	90 (7.9)	65 (5.2)d
Hypertension (yes) b , g	955 (58.2)	3077 (56.9)	599 (64.0)	670 (63.6)
Diabetes (yes) b	115 (6.7)	378 (6.6)	93 (10.3)	109 (10.7)
Alcohol intake
g/day c	2.4 [0.1–13.2]	3.9 [0.2–15.2]d
light b			778 (67.8)	842 (66.9)
moderate b			327 (28.5)	371 (29.5)
excessive b			42 (3.7)	46 (3.7)
Current smoking b	351 (20.4)	1345 (23.5)d	93 (8.1)	130 (10.3)
Lower limb disability (yes) b	601 (36.4)	1624 (29.4)d
Physical performance score c			9 [6–11]	9 [7–11]d
Handgrip strength (kg) c			29 [24–40]	33 [25–42]d
Dizziness (yes) b	616 (35.8)	1792 (31.1)d
Orthostatic hypotension (yes) b	203 (13.9)	656 (13.5)
eGFR (ml min −1  1.73 m −2 ) c	71.7 [61.7–81.5]	72.7 [62.2–82.7]d	70.8 [60.7–80.8]	72.2 [61.7–82.4]d
Selective β‐blockers (yes) b	161 (9.1)	623 (10.6)	235 (20.5)	247 (19.6)
Nonselective β‐blockers (yes) b	40 (2.3)	145 (2.5)	46 (4.0)	35 (2.8)
Antihypertensive use (yes) b	20 (1.1)	84 (1.4)	9 (0.8)	14 (1.1)
Diuretic use (yes) b	224 (12.7)	806 (13.7)	181 (15.8)	181 (14.4)
Benzodiazepine use (yes) b	236 (13.3)	648 (11.0)d	54 (4.7)	46 (3.7)
Antidepressant use (yes) b	35 (2.0)	128 (2.2)	57 (5.0)	38 (3.0)d

The numbers presented are based on the valid number of included fall cases and nonfallers.

^aPresented as mean (±standard deviation).

^bPresented as n (%).

^cPresented as median [interquartile range].

^dDifferences between fall cases and nonfallers within a study population with a P‐value <0.05.

^eFall history concerns falls in the last month for the Rotterdam Study and falls in the preceding year in B‐PROOF.

^fClinically relevant depressive symptoms were bases on Center for Epidemiological Studies Depression Scale scores ≥16^35,36, Hospital Anxiety and Depression Scale scores ≥9³⁷ or geriatric depression scale scores ≥5³⁸.

^gHypertension was defined as systolic blood pressure > 140 mmHg and/or diastolic blood pressure > 90 mmHg³¹.

MMSE = mini‐mental state examination, the estimated glomerular filtration rate (eGFR) is based on the chronic kidney disease epidemiology collaboration formula

Beta‐blocker use and fall risk

In both studies, current and past use of β‐blockers was not associated with fall risk, table 3. For current use – compared to nonuse – the combined HR was 0.97 [(95% confidence interval (CI) 0.88–1.06]. The use of selective β‐blockers was also not associated with fall risk, combined HR = 0.92 (95%CI 0.83–1.01). Use of nonselective β‐blockers was associated with an increased fall risk, combined HR = 1.22 (95%CI 1.01–1.48). Use of a lipophilic or nonlipophilic β‐blocker was not associated with fall risk, combined HR was 0.99 (95%CI 0.71–1.37), and HR = 0.99 (95%CI 0.88–1.09) respectively. In both studies, the hazard ratios were adjusted for age and sex, as the other considered covariates did not change the HR by more than 10%. As, in total, there were only four fall‐cases who used a β‐blocker with ISA capacity, the association between β‐blockers with and without ISA could not be investigated.

Table 3.

Association between β‐blocker use, β‐blocker characteristics – selectivity, lipid solubility, and CYP2D6 enzyme metabolism – and fall risk. Hazard ratios and 95% confidence intervals are presented

		Meta‐analysisb		Rotterdam Study			B‐PROOF
	No. of casesa	Model 1d	No. of casesc	Crude	Model 1d	No. of casesc	Crude	Model 1d
Beta‐blockers
Nonuse	1966	ref	1161	ref		805	ref
Past	363	1.07 (0.95– 1.20)	319	1.12 (0.98– 1.27)	1.07 (0.94– 1.21)	44	1.10 (0.81– 1.49)	1.07 (0.79– 1.46)
Current	588	0.97 (0.88– 1.06)	290	0.95 (0.84– 1.09)	0.96 (0.85– 1.10)	298	0.99 (0.87– 1.13)	0.98 (0.86– 1.12)
Selectivity
No β‐blockers use	2329	ref	1480	ref		849	ref
Selective β‐blockers	480	0.92 (0.83–1.01)	230	0.87 (0.76– 1.00)	0.90 (0.79– 1.04)	250	0.95 (0.82– 1.09)	0.94 (0.81– 1.08)
Nonselective β‐blockers	108	1.22 (1.01– 1.48)	60	1.24 (0.96– 1.61)	1.18 (0.91– 1.52)	48	1.30 (0.97– 1.74)	1.28 (0.96– 1.72)
Lipophilicity
No β‐blockers use	2329	ref	1480	ref		849	ref
Lipophilic β‐blockers	414	0.99 (0.88– 1.09)	173	1.01 (0.86– 1.18)	1.04 (0.88– 1.21)	241	0.94 (0.82– 1.09)	0.93 (0.81– 1.08)
Nonlipophilic β‐blockers	174	0.99 (0.71– 1.37)	117	0.84 (0.69– 1.01)	0.85 (0.70– 1.02)	57	1.23 (0.94– 1.60)	1.19 (0.91– 1.55)
CYP2D6 metabolism
No β‐blockers use	2329	ref	1480	ref		849	ref
2D6 metabolism e	334	0.98 (0.88– 1.11)	118	0.99 (0.82– 1.19)	1.00 (0.83– 1.21)	216	0.98 (0.85– 1.14)	0.97 (0.84– 1.13)
No 2D6 metabolism	254	0.94 (0.82– 1.07)	172	0.89 (0.76– 1.05)	0.92 (0.78– 1.07)	82	0.99 (0.80– 1.25)	0.98 (0.78– 1.23)

^aTotal number of fall cases per group of the Rotterdam Study and B‐PROOF combined;

^bMeta‐analyses results of the Rotterdam Study, and B‐PROOF;

^cNumber of fall cases per group;

^dModel 1: age, sex;

^eBeta‐blockers that are (partially) metabolized by the CYP2D6 enzyme, i.e., metoprolol, propranolol, nebivolol, carvedilol, timolol, and alprenolol (in the Rotterdam Study).

Dose response relation

In both studies, β‐blockers were used in relatively low dosages, the median number of prescribed DDDs was 0.50. No dose–response relation was observed. The combined analyses indicated that those using the median dose or less – compared to nonuse – had a fall risk of HR = 1.03 (95%CI 0.92–1.16). Those using a dose above the median had a fall risk of HR = 0.88 (95%CI 0.72–1.08). In addition, there was no significant linear trend for the dose categories, P = 0.159. Data of the individual studies are not shown.

Beta‐blocker use, fall risk and CYP2D6*4 genotype

In the Rotterdam Study, CYP2D6*4 allele had a frequency of 20%, and in B‐PROOF 22%. In both studies, the allele frequency was in Hardy–Weinberg equilibrium.

The association between use of β‐blockers, subjected to CYP2D6 enzyme metabolism or not, and fall risk, was stratified on CYP2D6*4 genotype. No significant associations were observed, Table 4.

Table 4.

Association between β‐blocker use, categorized by CYP2D6 enzyme metabolism and fall risk, stratified for CYP2D6*4 genotype. Hazard ratios and 95% confidence intervals (CI) are presented

		Meta‐analysesb		Rotterdam Study			B‐PROOF
CYP2D6*4	No. of casesa	Model 1d	No. of casesc	Crude	Model 1d	No. of casesc	Crude	Model 1d
Betablockers 2D6 Metabolism
No *4 carriers
Non use	1223	ref	757	ref		466	ref
2D6 metabolism e	176	0.92 (0.78–1.08)	64	1.03 (0.79–1.33)	1.02 (0.79–1.32)	112	0.87 (0.71–1.07)	0.86 (0.70–1.05)
No 2D6 metabolism	144	0.96 (0.81–1.14)	95	0.87 (0.71–1.08)	0.91 (0.73–1.12)	49	1.07 (0.80–1.43)	1.08 (0.81–1.45)
*4 carriers
Non use	745	ref	459	ref		286	ref
2D6 metabolism e	116	1.16 (0.95–1.42)	38	1.00 (0.72–1.40)	1.07 (0.77–1.49)	78	1.20 (0.94–1.55)	1.22 (0.95–1.56)
No 2D6 metabolism	81	0.98 (0.78–1.24)	54	1.02 (0.77–1.36)	0.99 (0.75–1.31)	27	1.03 (0.69–1.52)	0.97 (0.65–1.44)

^aTotal number of fall cases per group of the Rotterdam Study and B‐PROOF combined.

^bMeta‐analyses results of the Rotterdam Study, and B‐PROOF.

^cNumber of fall cases per group.

^dModel 1: age, sex.

^eBeta‐blockers that are (partially) metabolized by the CYP2D6 enzyme, i.e., metoprolol, propranolol, nebivolol, carvedilol, timolol and alprenolol (in the Rotterdam Study).

Only one of the meta‐analyses indicated significant heterogeneity between the studies, I² varied between 0–4%. The meta‐analyses of nonlipophilic β‐blockers indicated significant heterogeneity, p = 0.04 and I² = 4%.

Sensitivity analyses

Additional categorization of the nonselective β‐blocker group into nonselective, and nonselective highly lipophilic β‐blockers, did not result in materially different results. Use of nonselective β‐blockers – compared to nonuse – was nonsignificantly associated with an increased fall risk, combined HR = 1.22 (95%CI 0.97–1.53). Use of nonselective highly lipophilic β‐blockers was not significantly associated with fall risk, combined HR = 1.23 (95%CI 0.87–1.74). Likewise, subdividing lipophilic β‐blockers into medium and highly lipophilic β‐blockers did not show substantially different results, combined HR = 0.99 (95%CI 0.85–1.15), and HR = 1.22 (95%CI 0.86–1.71), respectively. In addition, no dose–response relation was observed for nonselective β‐blocker use (data not shown). The combined analyses indicated that those using the median dose (0.50 DDD) or less – compared to nonuse – had a fall risk of HR = 1.25 (95%CI 0.96–1.62). Those using a dose above the median had a fall risk of HR = 1.22 (95%CI 0.87–1.69). The results of the association for past and current use of selective and nonselective β‐blockers compared to nonuse are presented in Table S1. The results indicated no association with fall risk for past use of selective or nonselective β‐blockers. All these analyses were adjusted for age and sex.

Discussion

In two large older populations, the use of nonselective β‐blockers was associated with an increased fall risk. Use of selective, lipophilic or β‐blockers overall, was not associated with fall risk. Furthermore, we did not observe an association between β‐blocker use and fall risk across genotypes of CYP2D6.

To our knowledge, our study group is the first to evaluate the association between β‐blocker characteristics and fall risk. We observed an increased fall risk for current use of nonselective β‐blockers, although there was no dose–response relation. Furthermore, no association was observed for selective β‐blocker use or β‐blocker use overall. These findings might be explained by the receptor binding profile and accompanying systemic effects of nonselective β‐blockers. Nonselective β‐blockers, in addition to binding to β1‐receptors, bind to β2‐receptors and some also to α‐receptors. β1‐receptors are mainly located in the heart, while β2‐receptors are also present in the lungs, smooth muscle cells of the peripheral circulation, liver and in skeletal muscle cells 11, 12, 45. As a consequence, nonselective β‐blockers not only reduce heart rate and contractility, but also induce peripheral vasoconstriction, including in blood vessels towards and in skeletal muscle 11, 12. Contrarily, β‐ and α‐blockers also exhibit vasodilating properties 11, 12. In theory, β2‐antagonist may as well have a direct negative effect on skeletal muscle and might thereby be related to fall risk, as β2‐agonist are suggested to have a positive effect on muscle function 46, 47. Thus, nonselective β‐blockers may be related to fall risk by their broader range in effects and their potential negative effect on skeletal muscle.

Another aspect of selective and nonselective β‐blockers is that the indication for use can differ. Selective β‐blockers are mainly used for hypertension, although metoprolol, for example, is also used in patients with heart failure or those who with angina pectoris or a previous myocardial infarction. Nonselective β‐blockers, by contrast, are contraindicated for asthmatics and diabetics. Sotalol, a nonselective agent is used for arrhythmias and carvedilol is used for heart failure, but also for hypertension and angina pectoris 11, 12, 48. These potential indication differences of β‐blockers may be related to fall risk. Nevertheless, our sensitivity analyses did not indicate an association with fall risk for past use of selective or nonselective β‐blockers. This suggests that confounding by indication does not play a major role. If the association was spuriously caused by confounding by indication, we would expect a similar risk estimate in past users as in current users.

Within the B‐PROOF population we previously observed an association between nonselective β‐blocker use and fall risk, though then nonselective β‐blockers were slightly differently defined 8. Currently we also included in α‐ and β‐blockers, and excluded ocular administered β‐blockers.

In previous studies, overall use of β‐blockers has been associated with fall risk 9, 10, but not consistently 7. In the studies that reported an association, an increased fall risk was observed during initiation of use, which was thought to be due the increased risk of hypotension 9, 10. We investigated current use and not initiation of use, which may partly explain the discrepancy in results.

With respect to the lipophilicity of β‐blockers, some nonselective β‐blockers are highly lipophilic, such as carvedilol and propranolol 11, 12. Because lipophilic agents can cross the blood–brain barrier 11, 13, we hypothesized that their use was associated with more central adverse effects, including dizziness. Nevertheless, our sensitivity analyses investigating the use of β‐blockers, combining strong lipophilic and nonselective characteristics, in relation to fall risk did not result in a different association than for nonselective β‐blockers overall. Thus, our results do not confirm this hypothesis.

Regarding pharmacokinetic properties of β‐blockers, we hypothesized that users of β‐blockers that are subjected to 2D6 enzyme metabolism who carry a CYP2D6*4 variant are at increased risk for fall incidents, due to decreased metabolism and potentially increased drug concentrations. Previous studies indicated that the combination of metoprolol – a β‐blocker predominantly metabolized by CYP2D6 – use and a poor metabolizer phenotype – based on genotype – was associated with a lower clearance, longer half‐life 49, and with lower blood pressure and heart rate 17, 18. Although we were not able to investigate these specific endpoints, our results do not indicate that these clinical effects – whether they occurred or not – were translated into fall risk.

Our study has strengths and limitations. Its strength is the combination of two large, independent, community‐dwelling study populations, and thereby the possibility to investigate consistency of a potential signal (finding) across the two studies. Our study also has limitations, as the B‐PROOF study participants were included according to their homocysteine levels. However, we do not think that this inclusion criterion would have interfered with a mechanism underlying a potential association between β‐blocker use and falls. In the Rotterdam Study, fall incidents were differently assessed than in B‐PROOF, as serious fall incidents were gathered, and falls not leading to serious consequences were not included. This may lead to a different association if the underlying mechanism for β‐blocker‐related falls would differ between serious and less serious falls. However, we are not aware of different mechanisms, and the effect sizes were relatively similar across both studies. In addition, we investigated current use. Possibly, participants encountering side effects had already stopped using, switched to another β‐blocker or received lower doses. This may have resulted in underestimation of the association. Another limitation is the relatively low number of users carrying a CYP2D6*4 allele, consequently we clustered intermediate metabolizer phenotype (carriers of one *4 allele) with poor metabolizers (carriers of two *4 alleles). Despite the clustering, the numbers were too low to draw conclusion. Lastly, we do not have information on actual plasma levels of the β‐blockers.

Conclusion

Our study indicates an increased fall risk in older people during the use of nonselective β‐blockers, contrary to selective β‐blockers. In clinical practice, β‐blockers have been shown effective for a variety of cardiovascular indications. However, fall risk should be considered when prescribing a β‐blocker in this age group, and the pros and cons for β‐blockers classes should be taken into consideration.

Competing Interests

All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and the authors declare no potential conflicts of interest that are directly relevant to the content of this study.

We gratefully thank all study participants, and all dedicated coworkers who helped in the success of both studies. The Rotterdam Study is supported by the Erasmus MC and Erasmus University Rotterdam; the Netherlands Organization for Scientific Research (NWO); the Netherlands Organization for Health Research and Development (ZonMw); the Research Institute for Diseases in the Elderly (RIDE); the Netherlands Genomics Initiative (NGI); the Ministry of Education, Culture and Sciences; the Ministry of Health Welfare and Sports; the European Commission (DG XII); and the Municipality of Rotterdam. B‐PROOF is supported and funded by The Netherlands Organization for Health Research and Development (ZonMw, Grant 6130.0031 and 11‐31 010‐06), the Hague; unrestricted grant from NZO (Dutch Dairy Association), Zoetermeer; NCHA (Netherlands Consortium Healthy Aging) Leiden/Rotterdam; Ministry of Economic Affairs, Agriculture and Innovation (project KB‐15‐004‐003), the Hague; Wageningen University, Wageningen; VU University Medical Centre, Amsterdam; Erasmus Medical Center, Rotterdam. None of the funders had any role in design and conduct of this study; collection, management, analysis, and interpretation of the data; and review, or approval of the manuscript.

Contributors

Study concept and design: B.H.S., A.G.U. and N.v.d.V. Acquisition, analysis, or interpretation of data: all authors. Drafting of the manuscript: A.C.H., B.H.S., A.G.U. and N.v.d.V. Critical revision of the manuscript for important intellectual content: all authors. Statistical analysis and statistical expertise: A.C.H. and B.H.S. Obtained funding: B.H.S., A.G.U., and N.v.d.V.

Supporting information

Figure S1 Flowchart of the included study populations

Table S1 Association between selective and nonselective β‐blocker use for past and current use, and fall risk. Hazard ratios and 95% confidence intervals (CI) are presented

Ham, A. C. , van Dijk, S. C. , Swart, K. M. A. , Enneman, A. W. , van der Zwaluw, N. L. , Brouwer‐Brolsma, E. M. , van Schoor, N. M. , Zillikens, M. C. , Lips, P. , de Groot, L. C. P. G. M. , Hofman, A. , Witkamp, R. F. , Uitterlinden, A. G. , Stricker, B. H. , and van der Velde, N. (2017) Beta‐blocker use and fall risk in older individuals: Original results from two studies with meta‐analysis. Br J Clin Pharmacol, 83: 2292–2302. doi: 10.1111/bcp.13328.