Serotonin-Norepinephrine Reuptake Inhibitor Antidepressants and the Risk of Falls in Older People

Serotonin-Norepinephrine ReuptakeInhibitor Antidepressants and the Riskof Falls in Older PeopleCase-Control and Case-Series Analysis of a Large UK PrimaryCare Database

Jonathan Gribbin,1 Richard Hubbard,1 John Gladman,2 Chris Smith1 and Sarah Lewis1

1 Division of Epidemiology and Public Health, University of Nottingham, Nottingham, UK

2 Division of Rehabilitation and Ageing, University of Nottingham, Nottingham, UK

Abstract Background: Antidepressant medications have long been implicated as a

cause of falls in older people, but there are few data on the risk of falls

associated with exposure to serotonin-norepinephrine (noradrenaline) reup-

take inhibitors (SNRIs).

Objective: The aim of this study was to determine the role of SNRIs in older

people with a recorded fall in primary care using a case-control approach and

a self-controlled case-series analysis of data from The Health Improvement

Network (THIN) UK primary care database.

Methods: Cases were 9862 individuals aged >60 years with a first fall recorded

between 2003 and 2006. Up to six controls per case were matched by age, sex

and primary care practice. For the case-series analysis, we extended our case set

to those with a first fall recorded between 2001 and 2008. We estimated odds

ratios (ORs) for ‘ever’, ‘current’, ‘recent’, ‘previous’ or ‘never’ prescribed SNRIs

(and for the other main classes of antidepressants, for comparison) in cases

compared with matched controls, using conditional logistic regression. We also

examined the effect of the time interval from first prescription to first fall. In the

case-series analysis, we compared the rate of occurrence of first falls during

episodes of exposure to SNRIs with unexposed periods in our case group.

Results: There was an increased risk of current prescribing of SNRIs (ad-

justed OR 1.79; 95% CI 1.42, 2.25) in first fall cases compared with controls.

This was similar in magnitude to that seen with tricyclic antidepressants

(TCAs) and selective serotonin reuptake inhibitors (SSRIs). The increase in

risk was apparent within the first 28 days after first prescription. The effects

were also apparent in the self-controlled case-series analysis, although the

magnitudes of effect were slightly smaller; the incidence risk ratio for the

period 1–28 days after initiation of treatment compared with unexposed

periods was 1.49 (95% CI 1.15, 1.93).

ORIGINAL RESEARCH ARTICLEDrugs Aging 2011; 28 (11): 895-9021170-229X/11/0011-0895/$49.95/0

ª 2011 Adis Data Information BV. All rights reserved.

Conclusions: Treatment with SNRIs in older people may be associated with

an increased risk of falling. The falls risk profile of SNRIs appears to be

similar to that of SSRIs and TCAs.

Introduction

Older people are widely prescribed antide-pressants but are especially vulnerable to theirside effects. Antidepressant medications havelong been associated with an increased risk offalls and fracture in the elderly, and this increasein risk appears to be common to both the oldertricyclic antidepressants (TCAs) and the morerecent selective serotonin reuptake inhibitors(SSRIs).[1] It has been suggested that newer dual-action serotonin-norepinephrine (noradrenaline)reuptake inhibitors (SNRIs) may be safer interms of fracture risk[2] and risk of falling inresidential care and nursing homes,[3] but thereare no data on the risk of falls associated with theuse of these drugs in older people in the com-munity setting. We have therefore investigatedthe independent effect of SNRIs on the risk offalls in older people, and compared this risk tothat associated with other classes of antidepres-sants, using data from a large, UK, primary caredatabase and classical case-control analyses. Inthis between-person comparison, any tendencyfor preferential prescribing of SNRIs to those athigher risk of falls, and any increase in falls riskassociated with the underlying depression, maylead to an exaggeration of the effect. To exploreand minimize the potential for such bias, we alsoconducted a within-person (self-controlled) anal-ysis, using the case-series method.[4]

Methods

The Health Improvement Network (THIN) isa longitudinal primary care database, containingthe diagnostic and prescribing data recordedby primary care physicians, as part of routineclinical care, on over 6 million patients from over400 general practices throughout the UK, all ofwhich use the Vision Practice Management sys-tem.[5] THIN includes information from hospital

referrals and emergency admissions, and the dateand cause of death. The database has a patientdemographic similar to that of the general pop-ulation in the UK.

Case-Control Analysis

For the case-control analysis, we used datafrom the 386 primary care practices that con-tributed data for the entire period 2003–6 andincluded only those patients aged ‡60 yearsduring this study period. From this study pop-ulation, we defined cases as those who experi-enced their first recorded fall during the studyperiod (the index date). We defined fall eventsusing Read Codes, which comprise the compre-hensive set of symptoms and diagnoses availableto primary care practitioners in the UK forrecording the reason for a consultation or otherevent. Out of the wide range of Read Codesavailable, 80% of events relating to falls arerecorded using the term ‘Fall – accidental.’ Thisterm and a further seven account for 99% offall events recorded in THIN. We defined a firstfall as the first event recorded using a Read Codecorresponding to a fall. A list of the codes isavailable on request from the authors.

For each case, we randomly selected up to sixgeneral population controls, matched on age, sexand primary care practice, who had no recordedfalls and were contributing data at the date ofthe index event for the case, as described in ourprevious comparison of mortality between thesecase and control groups.[6] Cases and controlswere required to have at least 12 months ofrecorded history prior to the index date. Wedetermined that this case-control analysis wouldhave over 99% power to detect an odds ratio(OR) of 1.5 or greater for an exposure occurringin 2% of controls.

Our main exposures of interest were prescrip-tions for SNRIs issued prior to the index date,

896 Gribbin et al.

ª 2011 Adis Data Information BV. All rights reserved. Drugs Aging 2011; 28 (11)

and these drugs were grouped according to theirclassification in the British National Formulary.[7]

We also examined the effect of prescriptions forother classes of antidepressant to validate ourdata by establishing whether the associationswere consistent with those in the literature,[1,8,9]

and to compare the risks associated with SNRIswith those for other antidepressant classes.

Firstly, we classified exposure to each classof antidepressant in terms of the elapsed timebetween the index event and the final precedingprescription, as follows: ‘ever prescribed’, ‘cur-rent’ (last prescription within 60 days of indexevent), ‘recent’ (60–120 days), ‘previous’ (morethan 120 days) or ‘never prescribed’. Secondly,we explored the extent to which an increased riskof falling might be associated with the initiationof prescribing. We did this by categorizing firstfalls in terms of the elapsed time from the timeof the first prescription in categories of 0 days(day of prescription), 1–28 days, 29–56 days and>56 days to assess the longer term effects. Weseparated the day of prescription (day 0) from theremaining days of the first 4 weeks of antidepres-sant prescription (days 1–28) in order to examinethe risk in the first month without the ascertain-ment bias that would result from falls recorded onthe day of the antidepressant prescription.

We used conditional logistic regression toestimate ORs for each of our exposures in theunivariate analysis. We then examined the im-pact of a number of potentially confoundingvariables, including coronary heart disease, dia-betes mellitus, cardiovascular disease, other co-morbidity and prescription of antipsychotics,hypnotics-anxiolytics, diuretics, digoxin or anti-arrhythmics, in a series of bivariate models. Ourmeasure of general co-morbidity was an adapta-tion of the Charlson Comorbidity Index used as apredictor of 1-year mortality based on previousdiagnoses of 17 diseases, including myocardialinfarction, congestive heart failure, peripheralvascular disease, diabetes and cerebrovasculardisease.[10] If any of these potentially confound-ing variables led to a ‡10% change in the OR forSNRI exposure, it was retained in the multi-variate model. We tested for interaction with ageusing the likelihood ratio test. We used a similar

approach to explore the relationship betweenfirst fall and first prescription.

Self-Controlled Case-Series Analysis

In the case-series method, the incidence of fallsin defined periods of time after exposure toSNRIs was compared with the incidence in otherbaseline periods of time for the same individual.This analysis used data from cases only, whichwere defined as for the case-control analysis, butwe extracted further data from THIN to extendour case set to those with a first recordedfall between 2001 and 2008 to ensure adequatepower. With an average follow-up of 6 years, werequired 1900 exposed cases to provide 90%power to detect an OR of 1.7 for an increase inrisk in the first 28 days of exposure.

We divided person-time into different timeperiods after the start of treatment (high-riskperiods) as follows: day 0 (day of the start oftreatment), days 1–28, days 29–56 and day 57 tothe end of the treatment episode. Consecutiveprescriptions separated by not more than 60 dayswere treated as belonging to a single episode, withthe endpoint of each episode being defined as60 days after the last prescription in that episode.All remaining person-time was used as the base-line (unexposed) comparison period.

We estimated the relative incidence for theserisk periods using conditional Poisson regression,adjusting for age in 1-year age bands. We exam-ined the confounders used in our case-controlmodel, retaining them in the case-series analysiswhere they changed the relative risk by ‡10%.We performed a sensitivity test to assess theassumption underlying the case-series methodthat the occurrence of an event does not alter theprobability of subsequent exposure.[4] To do this,we amended the standard analysis to include anadditional observation period correspondingto the 30 days prior to the start of an episode. Wecompared the incidence rate ratios for thestandard analysis with the respective risk periodsin the sensitivity analysis, on the basis that a largechange in incidence rate ratios would be evidencethat events influence the probability of a sub-sequent exposure.[4]

SNRIs and Risk of Falls in Older People 897


The study protocol was reviewed and approvedby the Nottingham Research Ethics Committee.

We used STATA, version 9 (Statacorp, Col-lege Station, TX, USA) for all statistical analysesand likelihood ratio tests to test for heterogeneityacross categories.

Results

We identified 9682 people aged >60 yearswho experienced a first fall, and 52 100 matchedcontrols (table I). The mean age of cases was77.5 years, and 76.4 years for controls. Thirty-two percent of cases and controls were male.

The results of our analyses of ever (previous/recent/current)/never prescribing are shown intable II. Prescribing of SNRIs was uncommoncompared with prescribing of TCAs and SSRIs;of the controls, 806 (2%) had ever been prescribedan SNRI, and most of these (768) were prescribedvenlafaxine, with small numbers having beenprescribed reboxetine (41) or duloxetine (19).Cases who had experienced a first fall were morelikely than controls to have ever been prescribedan SNRI (3%), and the unadjusted OR associatedwith having ever been prescribed an SNRI was1.77 (95% CI 1.53, 2.05). The risk of first fall washighest in those currently prescribed an SNRI,

with a 2.5-fold increase in risk compared withthose who had never been prescribed an SNRI.Of the potential confounders we evaluatedfor inclusion in bivariate models, prescribing ofantipsychotics and hypnotic-anxiolytic drugsaltered the effect by ‡10%. In our final model,after adjustment for antipsychotic and hypnotic-anxiolytic prescribing, the effect of current use ofSNRIs was reduced but still demonstrated analmost 80% increase in risk (adjusted OR 1.79;95% CI 1.42, 2.25). The magnitudes and signi-ficance of effects were very similar when pre-scriptions were limited specifically to venlafaxine.

We found evidence of an interaction with agefor TCAs (p = 0.003), for which the size of theeffect (i.e. risk of first fall) increased with age. ForSNRIs and SSRIs, we found no evidence ofinteraction with age (data not shown).

The risk of first fall was apparent withinthe first 28 days following first prescription of anSNRI (adjusted OR 3.42; 95% CI 0.96, 12.24)and was maintained in the subsequent 28 days(adjusted OR 4.40; 95%CI 1.46, 13.27) [table III].

When we validated our analysis by searchingfor the anticipated effects of other antidepres-sants, we found an increased risk of first fallwith ever having been prescribed either a TCA(unadjusted OR 1.64) or an SSRI (unadjustedOR 1.88) [table II]; these values were very similarin size to the effects on hip fracture previouslydescribed for this cohort.[11] The adjusted OR forfirst fall with current exposure to SSRIs was 2.04(95% CI 1.86, 2.24), which was slightly greaterthan that for TCAs [1.61 (95%CI 1.46, 1.76)] andsimilar to other findings for antidepressants.[9]

Both TCAs and SSRIs were associated with anacute increase in risk in the first few weeks afterthe first prescription, which was more marked forTCAs than for SSRIs. The effect of SNRIs on thefirst fall in our analyses was very similar to thatseen for SSRIs and for TCAs.

When the data were extended to cover theperiod 2001–8, there were 1916 fall cases exposedto SNRIs.When we conducted the self-controlledcase-series analysis (table IV), comparing the riskof falls in periods of exposure with baseline (un-exposed) periods, the risk was found to be signi-ficantly increased in the period 1–28 days after

Table I. Characteristics of cases and controls in The Health

Improvement Network (THIN) 2003–6a

Characteristic Cases Controls

Participants (n) 9682 52 100

Age (y; mean) 77.5 76.4

Sex

females 6602 (68) 35 573 (68)

males 3080 (32) 16 527 (32)

CHD (ever) 6974 (72) 34 051 (65)

Diabetes mellitus or CVD (ever) 5425 (56) 23 559 (45)

Current prescription

antipsychotic 509 (5) 1 371 (3)

hypnotic/anxiolytic 1208 (12) 3 744 (7)

diuretic 3657 (38) 16 459 (32)

digoxin 594 (6) 2 022 (4)

type 1a anti-arrhythmic 0 (0) 0 (0)

a Values are n (%) unless otherwise stated.

CHD = coronary heart disease; CVD = cardiovascular disease.

898 Gribbin et al.


commencement of an episode of SNRI treatment(although smaller than that seen in the case-control analyses) and in the period from day 57 tothe end of treatment. Adjustment for the con-founders used in the case-control model did notalter the effect by ‡10%. A sensitivity analysis totest the possible influence of fall events on risk ofsubsequent exposure resulted in little change inthe size of the effect. The results were also verysimilar when exposure was restricted to venla-faxine treatment alone.

Discussion

Findings

In this large, case-control study of older peo-ple that used prospectively collected data, wefound that the risk of first fall was increased al-most 2-fold in people currently prescribed anSNRI, and this increase in risk was apparent inthe first month after the first prescription. Thesizes of effect were similar to those seen for SSRIs

and TCAs. In our within-person analysis, theeffect of SNRIs was smaller, suggesting thatour case-control study findings may have beenoverestimates.

Study Strengths and Weaknesses

The main strengths of our study include thelarge number of cases and the fact that our datawere collected prospectively, which means thatrecall of medication exposures and of falls is not asource of bias. In the UK, antidepressant medi-cations are available only on prescription, andprevious studies have shown these to be welldocumented where there is a computerized sys-tem.[12] Therefore, in settings in which practicescontribute to THIN, the data provide a reliablemeasure of prescribing. We differentiated medi-cations according to standard classifications inthe British National Formulary.[7] We validated thedata and the analysis using other classes of anti-depressants and these exhibited the expected pat-tern of risk, based on evidence from other studies.

Table II. Association between first recorded fall and prescribing of antidepressants in The Health Improvement Network (THIN) 2003–6

Exposure Cases [n (%)]

(n =9682)Controls [n (%)]

(n= 52 100)OR (95% CI) Adjusteda OR (95% CI)

SNRI

Never prescribed 9426 (97) 51 294 (98) 1.00 1.00

Ever prescribed 256 (3) 806 (2) 1.77 (1.53, 2.05) 1.30 (1.12, 1.51)

previously 135 (1) 516 (1) 1.47 (1.21, 1.78) 1.09 (0.90, 1.33)

recently 6 (0) 34 (0) 0.98 (0.41, 2.35) 0.79 (0.33, 1.90)

currently 115 (1) 256 (0) 2.49 (1.99, 3.12) 1.79 (1.42, 2.25)

SSRI


Ever prescribed 1956 (20) 6 350 (12) 1.88 (1.77, 1.99) 1.55 (1.46, 1.65)

previously 1081 (11) 4 208 (8) 1.57 (1.46, 1.69) 1.31 (1.21, 1.41)

recently 111 (1) 297 (1) 2.24 (1.79, 2.80) 1.85 (1.48, 2.32)

currently 764 (8) 1 845 (4) 2.50 (2.29, 2.74) 2.04 (1.86, 2.24)

TCA


Ever prescribed 2615 (27) 9 891 (19) 1.64 (1.56, 1.72) 1.38 (1.30, 1.46)

previously 1811 (19) 7 313 (14) 1.54 (1.45, 1.63) 1.30 (1.23, 1.39)

recently 106 (1) 390 (1) 1.70 (1.36, 2.11) 1.44 (1.16, 1.80)

currently 698 (7) 2 188 (4) 1.96 (1.79, 2.14) 1.61 (1.46, 1.76)

a Adjusted for prescribing of antipsychotics and hypnotics-anxiolytics.

OR = odds ratio; SNRI = serotonin-norepinephrine (noradrenaline) reuptake inhibitor; SSRI = selective serotonin reuptake inhibitor;

TCA = tricyclic antidepressant.



The main potential weaknesses of our studyare the validity of the recording of fall events, andthe incomplete control of confounding. Our pre-vious study suggested that falls recorded in pri-mary care are a subset of the falls self-reported insurveys.[6] We do not have data on which self-reported falls are more likely to be recordedin primary care, but our earlier study foundthat patients with recorded falls are an importantgroup who experience increased mortality com-pared with non-fallers (2-fold increase for re-current fallers, and a more than 5-fold increasefor recurrent fallers aged 60–74 years), whichsuggests that recorded falls are more likely to bethose for which medical attention was required.[6]

Nevertheless, incomplete recording of falls raisesthe possibility of differential ascertainment offalls for patients receiving antidepressant med-ication. Such a bias would have led to over-estimated ORs. However, we excluded the day ofthe prescription from our exposure time, whichwould have avoided ascertainment bias of fallsrecorded at the time of initial prescribing. More-

over, the sizes of effect that we found for TCAsand SSRIs are very similar to those previouslyreported in this dataset for hip fracture, which isrelatively unlikely to be under-recorded.[11]

We explored the possibility that our resultsare affected by confounding by co-morbidity inseveral ways. In our case-control analysis, wecontrolled for a number of potential confoundersand adjusted for prescription of antipsychoticsand hypnotic-anxiolytics in our final model,which were the only potential confounders to re-duce the OR for the effect of SNRIs. We alsoexplored the extent to which the effect was attri-butable to characteristics of the patients ratherthan to prescribing, by comparing the risk forpeople currently and previously prescribedSNRIs; this analysis revealed much strongereffects for current prescribing. Finally, we used acase-series analysis, a method that controls forfactors that vary between individuals, such asfrailty and severity of depression. The smaller sizeof effect in the case-series analysis suggests thatthe case-control analysis may have overestimated

Table III. Association between first recorded fall and first prescription of antidepressant in The Health Improvement Network (THIN) 2003–6

First prescription (days elapsed before first fall) Cases [n (%)]

(n =9682)Controls [n (%)]

(n= 52100)OR (95% CI) Adjusteda OR (95% CI)

SNRI

No prescriptions prior to index date 9426 (97) 51 294 (98) 1.00 1.00

0 days 0 (0) 1 (0)

1–28 days 5 (0.05) 5 (0.01) 5.10 (1.45, 17.94) 3.42 (0.96, 12.24)

29–56 days 7 (0.07) 6 (0.01) 6.91 (2.32, 20.58) 4.40 (1.46, 13.27)

>56 days 244 (3) 794 (2) 1.71 (1.48, 1.98) 1.26 (1.09, 1.47)

SSRI


0 days 6 (0.06) 0 (0)

1–28 days 31 (0.32) 67 (0.13) 2.74 (1.79, 4.22) 2.34 (1.52, 3.62)

29–56 days 36 (0.37) 54 (0.10) 3.62 (2.35, 5.58) 3.09 (2.00, 4.78)

>56 days 1883 (19) 6 229 (12) 1.85 (1.74, 1.96) 1.52 (1.43, 1.62)

TCA


0 days 8 (0.08) 7 (0.01) 7.38 (2.66, 20.45) 7.12 (2.56, 19.82)

1–28 days 29 (0.30) 52 (0.10) 3.51 (2.22, 5.57) 3.18 (2.00, 5.07)

29–56 days 21 (0.22) 65 (0.12) 1.91 (1.16, 3.15) 1.65 (0.99, 2.73)

>56 days 2557 (26) 9 767 (19) 1.62 (1.54, 1.71) 1.36 (1.29, 1.44)

a Adjusted for prescribing of antipsychotics and hypnotics-anxiolytics.

OR = odds ratio; SNRI = serotonin-norepinephrine (noradrenaline) reuptake inhibitor; SSRI = selective serotonin reuptake inhibitor; TCA =tricyclic antidepressant.

900 Gribbin et al.


the true effect, because of bias. Nevertheless, thefinding of increased fall risk in the first 28 daysafter the start of treatment with SNRIs in thecase-series analysis suggests an independent ad-verse effect of this drug on fall risk.

We could not judge the extent, if any, of non-compliance to medication, and we relied on thetiming of prescriptions to approximate periods ofexposure and non-exposure. However, we alsonote that misclassifications in exposure wouldresult in underestimation of a true effect.

Other Studies

Several recent meta-analyses have demonstra-ted that antidepressant prescribing is associatedwith an increased incidence of falling, based onstudies conducted in a variety of settings.[1,9,13]

However, these studies did not evaluate SNRIsand falls in community-dwelling older people. In alarge, case-control study in Denmark (n= 498 617),Vestergaard et al.[2] investigated the effect ofthe full spectrum of antidepressant groups onfracture risk and found significant associationswith fracture at specific sites for SNRIs but noconsistent increase in risk. A study of residentialcare and nursing homes in Sweden (n = 3604)found no significant increase in risk of falls withSNRIs, although the investigators did observe aneffect with SSRIs.[3] Both of these studies sug-gested that SNRIs may be a safer alternative forolder people than TCAs or SSRIs, which appearto be associated with both falls and frac-ture.[1,3,9,11,14,15] Our study is the first to examinethe effect of SNRIs on risk of falls in the com-munity and suggests there is an increase in risksimilar to that for SSRIs, and an increased riskshortly after initiation of therapy. A recent review

observed that the mechanisms by which antide-pressants increase risk of falls and related frac-tures are complex and may include orthostatichypotension, arrhythmias, sedation, insomnia,movement disorders and confusion.[16] The specificmechanism bywhich SNRIs cause falls is not clear,but venlafaxine, which is the most commonly pre-scribed SNRI, has adverse effects in common withSSRIs and TCAs (e.g. drowsiness, confusion,insomnia, movement disorders, visual disturbanceand hypotension), each of which are implicated ascontributing to falls in older people.[16]

The evidence for the relative efficacy of SNRIscompared with other antidepressants remainslimited and inconsistent.[17-19] Nevertheless, SNRIshave been suggested by some to be the antide-pressant of choice for the elderly because of theirlow potential for drug interactions and possiblefavourable effect on pain associated with depres-sion,[18-20] although concerns about cardiotoxicityand toxicity in overdose have led the NationalInstitute for Health and Clinical Excellence(NICE) to recommend that these drugs be used assecond-line treatments. Our results suggest thatclinicians initiating prescribing of SNRIs shouldalso be alert to the increased risk of falls.

Conclusions

Our study provides evidence that SNRIs areassociated with an increased risk of first falls inolder people that is similar to that associated withother classes of antidepressants.

Acknowledgements

No sources of funding were used to conduct this study orprepare the manuscript. Jonathan Gribbin received a grant

Table IV. Association between SNRI antidepressant exposure (any treatment episode) and first fall: case-series analysis in The Health

Improvement Network (THIN) 2001–8

Period of exposure First fall (n) Follow-up time (pyrs) Incidence rate ratio (95% CI)

Unexposed (pre-/post-/between episodes) 1308 8960.7 1.00

Day 0 of episode 17 11.7 10.03 (6.15, 16.36)

Day 1–28 of episode 71 329.9 1.49 (1.15, 1.93)

Day 29–56 of episode 56 330.5 1.18 (0.88, 1.57)

Day 57 to end of episode 464 2669.4 1.29 (1.08, 1.53)

pyrs =patient-years; SNRI = serotonin-norepinephrine (noradrenaline) reuptake inhibitor.



from East Midlands Strategic Health Authority to cover stu-dent fees. The authors have no conflicts of interest that aredirectly relevant to the content of this study.

References1. Leipzig RM, Cumming RG, Tinetti ME. Drugs and falls in

older people: a systematic review and meta-analysis. II:cardiac and analgesic drugs. J AmGeriatr Soc 1999 Jan; 47(1): 40-50

2. Vestergaard P, Rejnmark L, Mosekilde L, et al. Selectiveserotonin reuptake inhibitors and other antidepressantsand risk of fracture. Calcif Tissue Int 2008 Feb; 82 (2):92-101

3. Kallin K, Gustafson Y, Sandman PO, et al. Drugs and fallsin older people in geriatric care settings. Aging Clin ExpRes 2004 Aug; 16 (4): 270-6

4. Whitaker HJ, Farrington CP, Spiessens B, et al. The self-controlled case series method. Stat Med 2006; 25 (10):1768-97

5. INPS. Vision system. 2011 [online]. Available from URL:http://csdmruk.cegedim.com/our-data/statistics.html [Ac-cessed 2011 Sep 12]

6. Gribbin J, Hubbard R, Smith C, et al. Incidence and mor-tality of falls amongst older people in primary care in theUnited Kingdom. QJM 2009 Jul 1; 102 (7): 477-83

7. British Medical Association, Britain RPSoG. British na-tional formulary. London: BMJ Publishing Group, 2008

8. Hartikainen S, Lonnroos E, Louhivuori K. Medication as arisk factor for falls: critical systematic review. J Gerontol ABiol Sci Med Sci 2007 Oct; 62 (10): 1172-81

9. Woolcott JC, Richardson KJ, Wiens MO, et al. Meta-analysis of the impact of 9 medication classes on falls inelderly persons. Arch Intern Med 2009 Nov 23; 169 (21):1952-60

10. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical co-morbidity index for use with ICD-9-CM administrativedatabases. J Clin Epidemiol 1992 Jun; 45 (6): 613-9

11. Hubbard R, Farrington P, Smith C, et al. Exposure to tri-cyclic and selective serotonin reuptake inhibitor antide-

pressants and the risk of hip fracture. Am J Epidemiol 2003Jul 1; 158 (1): 77-84

12. Thiru K, Hassey A, Sullivan F, et al. Systematic review ofscope and quality of electronic patient record data in pri-mary care. BMJ 2003 May 17; 326 (7398): 1070-4

13. Sterke CS, Verhagen AP, van Beeck EF, et al. The influenceof drug use on fall incidents among nursing home residents:a systematic review. Int Psychogeriatr 2008 Oct; 20 (5):890-910

14. Ziere G, Dieleman JP, van der Cammen TJ, et al. Selectiveserotonin reuptake inhibiting antidepressants are asso-ciated with an increased risk of nonvertebral fractures.J Clin Psychopharmacol 2008 Aug; 28 (4): 411-7

15. Richards JB, Papaioannou A, Adachi JD, et al. Effect ofselective serotonin reuptake inhibitors on the risk of frac-ture. Arch Intern Med 2007 Jan 22; 167 (2): 188-94

16. Darowski A, Chambers SA, Chambers DJ, et al. Antidepres-sants and falls in the elderly.Drugs Aging 2009; 26 (5): 381-94

17. Mottram PG, Wilson K, Strobl J. Antidepressants for de-pressed elderly. Cochrane Database Syst Rev 2006 Jan 25;(1): CD003491

18. Mukai Y, Tampi RR, Mukai Y, et al. Treatment of depres-sion in the elderly: a review of the recent literature on theefficacy of single- versus dual-action antidepressants. ClinTher 2009 May; 31 (5): 945-61

19. Papakostas GI, Thase ME, Fava M, et al. Are antidepres-sant drugs that combine serotonergic and noradrenergicmechanisms of action more effective than the selectiveserotonin reuptake inhibitors in treating major depressivedisorder? A meta-analysis of studies of newer agents. BiolPsychiatry 2007 Dec 1; 62 (11): 1217-27

20. Sussman N. SNRIs versus SSRIs: mechanisms of action intreating depression and painful physical symptoms. PrimCare Companion J Clin Psychiatry 2003; 5 Suppl. 7: 19-26

Correspondence: Jonathan Gribbin, Division of Epidemiol-ogy and Public Health, Clinical Sciences Building, CityHospital, Hucknall Road, Nottingham NG5 1PB, UK.E-mail: [email protected]

902 Gribbin et al.


Documents

Serotonin-Norepinephrine Reuptake Inhibitor Antidepressants and the Risk of Falls in Older People