European Neuropsychopharmacology (2013) 23, 872–878

0924-977X/$ - see frhttp://dx.doi.org/1

nCorresponding auE-mail address: j

www.elsevier.com/locate/euroneuro

Schizophrenia, antipsychotics and risk of hipfracture: A population-based analysis

Holger J. Sørensena,c, Signe O.W. Jensenb, Jimmi Nielsena,n

aCentre for Schizophrenia, Aalborg Psychiatric Hospital, Aarhus University Hospital, Brandevej 5,9220 Aalborg Øst, DenmarkbUnit for Psychiatric Research, Aalborg Psychiatric Hospital, Aarhus University Hospital,Mølleparkvej 10, P.O. Box 210, DK-9100 Aalborg, DenmarkcMental Health Center Copenhagen, Bispebjerg Bakke 13A, 2400 Copenhagen NV, Denmark

Received 14 November 2012; received in revised form 24 March 2013; accepted 2 April 2013

KEYWORDSFirst-generation anti-psychotics (FGA);Second-generationantipsychotics (SGA);Hyperprolactinemia;Dose–response rela-tionship;Anticholinergics;Schizophrenia

ont matter & 20130.1016/j.euroneur

thor. Tel.: +45 257in@rn.dk (J. Niels

AbstractIn a nationwide study using linkage of Danish hospital registers we examined predictors of hipfracture (ICD-10: S72) in 15,431 patients with schizophrenia (ICD-10: F20 or ICD-8: 295) and3,807,597 population controls. Shorter education, disability pension, lifetime alcohol abuse, somaticco-morbidity, antipsychotics (IRR=1.19; 95% CI 1.15–1.24), antidepressant (IRR=1.18; 95% CI 1.16–1.20), anticholinergics (IRR=1.29; 95% CI 1.22–1.36), benzodiazepines (IRR=1.06; 95% CI 1.04–1.08)and corticosteroids (IRR=1.44; 95% CI 1.36–1.53) were significant predictors. In 556 persons withschizophrenia and hip fracture (matched to 1:3 to schizophrenia controls without hip fracture),antipsychotic polypharmacy predicted hip fracture. Analyses among antipsychotic monotherapypatients showed no differential effect of individual antipsychotics. A dose–response relationship ofhip fracture and lifetime antipsychotics consumption was found (IRR=1.13 95% CI 1.07–1.19) andboth prolactin-increasing and non-prolactin-increasing antipsychotics contributed to the effect. Inconclusion, several factors, including complex psychopharmacological treatment, contribute in theprediction of hip fracture in large populations. Preventive strategies should focus attention toseverely ill patients with high likelihood of a receiving complex psychopharmacologic treatment andhigh doses of antipsychotics.& 2013 Published by Elsevier B.V.

1. Introduction

Hip fracture is associated with significant health care costs andconsiderable morbidity and mortality (Cumming et al., 1997;Cummings and Melton, 2002). Relatively little is known aboutthe contribution of a diagnosis of schizophrenia to the risk of

Published by Elsevier B.V.o.2013.04.002

79926.en).

hip fracture throughout lifetime. The illness, comorbid condi-tions, lifestyle, and psychopharmacological treatment may allincrease the risk. Decreased bone mineral density has beenfound in early in the course of schizophrenia (Abraham et al.,2003; Partti et al., 2010) and possible mediators includehyperprolactinemia and hypercortisolemia (Halbreich andPalter, 1996). Antipsychotic drugs may in the short term, andpossibly through alpha-1 noradrenergic and histamineantagonistic effects, increase the risk of falling. A number ofstudies have furthermore addressed the long-term effects of

873Schizophrenia, antipsychotics and risk of hip fracture

antipsychotics in relation to risk of hip fracture and possiblemechanisms behind the association (Howard et al., 2007;Hugenholtz et al., 2005; Pouwels et al., 2009; Vestergaardet al., 2006). A large UK based study found that schizophreniaand prolactin-increasing antipsychotics were associated withincreased risk of hip fracture in univariate analyses but in lightof a strong association with prolactin-increasing antipsycho-tics, the effect of schizophrenia did not remain significant(Howard et al., 2007). A meta-analysis of studies of fracturesin elderly people found a 1.6 fold elevated risk associated withany antipsychotic drug (Takkouche et al., 2007). In a samplewith schizophrenia, current and prior use of antipsychoticsincreased the risk of hip fracture but no significant associationswith dose or degree of blockade of the alpha-1 receptor orhistamine-1 receptor were found (Hugenholtz et al., 2005).Bolton et al. found osteoporotic fracture to be associated withschizophrenia, but observed no independent effect of anti-psychotics (Bolton et al., 2008). Pouwels et al. found a 1.76fold increased risk of hip fracture in elderly people treatedwith conventional but not atypical antipsychotics (Pouwelset al., 2009). In light of these previous findings, we initiatedthe current nationwide study to further examine predictors ofhip fracture in a large sample of patients with schizophrenia.

2. Experimental procedures

2.1. Registers

All psychiatric admissions have been registered in a nationwideregister, the Danish Psychiatric Central Research Register (DPCRR)from April 01, 1970 and onwards (Mors et al., 2011; Munk-Jorgensenand Mortensen, 1997). The ICD-8 was used until 1994 then replacedby ICD-10. DPCRR has been used extensively for schizophreniaresearch and the validity of the diagnosis is high (Uggerby et al.,2013). The Danish Data Protection Agency, National Board of Healthand Statistics Denmark approved use of data for this study. Datawere obtained by linking Danish population-based registers usingthe unique personal identification number (CPR-number), thusensuring accurate linkage of information between registers. Datafrom the national prescription database was linked with the DanishNational Patient Registry and the Danish Psychiatric CentralResearch Register. Prescription data for the included subjects wereobtained from the national prescription database from January 01,1995 until December 31, 2009. The prescription database containsinformation about number of sold defined daily doses (DDD). DDDfor each drug is assigned by the World Health Organization (WHO)and has been used to compare exposure to psychotropics in similarstudies (Nielsen et al., 2009; Nielsen and Meyer, 2012).

2.2. Samples

Exposed subjects for the analyses were individuals diagnosed withICD-8 295 or ICD-10 F20.x schizophrenia diagnosis before January01, 1995 and who were alive by January 01, 2000. The outcome waship fracture (ICD-10 code: S72). We excluded individuals with ahistory of hip fracture before January 01, 2001 and obtained dataon incident hip fracture from January 01, 2001 until December 31,2009. This was done to ensure at least 5 years of prescriptionhistory before the date of hip fracture.

The study population consisted of 15,431 patients with schizophre-nia and 3,807,597 population controls. Initial analyses examined thedemographic characteristics of patients with schizophrenia and con-trols. In a matched sample of schizophrenia patients (n=2224) wefurther estimated the effect of current exposure to antipsychotic

(within 4 months of the date of a diagnosis of hip fracture) and thelong-term effect of dosage of cumulative consumption antipsychotics(antipsychotics were subdivided into prolactin-increasing antipsychotics(PRL-AP) and non-prolactin-increasing antipsychotics (non-PRL-AP).

2.3. Medications

Antipsychotics were first divided into first-generation antipsychotics(FGA) and second generation antipsychotics (SGA) and secondly dividedinto prolactin elevating antipsychotics (risperidone, amisulpride, high-potency and mid-potency FGAs) and all other antipsychotics. Antipsy-chotics were further divided in low-, mid- and high-potency FGAs(definitions used from a previous study (Nielsen and Meyer, 2012)). Thelow potency FGA group consisted of clorpromazine, levopromazine,chlorprotixene, and pipamperone. The medium potency FGA groupsconsisted of zuclopenthixol, prochlorperazine, and perphenazine. Thehigh potency FGA group consisted of haloperidol, pimozide, flu-penthixol, and flupenazine. The SGA group consisted of amisulpiride,aripripazole, clozapine, olanzapine, quetiapine, risperidone, sertindole,sulpiride, and ziprasidone. The anticholinergic group consisted ofatropine, benztropine, biperidine, orphenadrine, and procyclidine.The group of antidepressants consisted of amitriptyline, clomipramine,doxepin, imipramine, and nortriptyline, the group of SSRI's andthe group of SNRI's. The group of corticosteroid consisted of ATC codeH02AB.

In order to estimate the effect of current psychopharmacologicaltreatment, medication status was based on prescriptions pickedup from the pharmacy within the last 4 months before index date.Four months were chosen because this includes patients receivingone tablet per day with the largest pack size (100 tablets).

For further analyses, we calculated the average monthly DDDs ofPRL-AP and non-PRL-AP as lifetime use based on the history ofprescriptions picked up from the pharmacy at any time betweenJanuary 01, 1995 and the index date.

2.4. Other variables

From Statistics Denmark we obtained information about educationalbackground and early retirement pension of cohort members andthese variables were coded as described elsewhere (Uggerby et al.,2011). From the Danish National Patient Register we obtainedinformation about physical illness and used a composite physicaldisease score (CPDS) measuring the total load of physical morbidityin the Danish population (Ostergaard et al., 2013). The CPDS is aninteger score with a theoretical range from 0 to 17 calculated byadding each physical disease (a patient diagnosed with psoriasis wouldreceive a CPDS of 1 while a patient diagnosed with Parkinson's diseaseknown to be quite strongly associated with falling and hip fracture(Vestergaard et al., 2007) would also receive a CPDS of 1). We used apragmatic measure of the CPDS (score ≥1 yes/no) to adjust forphysical morbidity. Adjustment for alcohol misuse was done using theDanish National Patient Registry and the national prescription data-base (individuals with a history of alcohol-related diagnoses: ICD-10code F10, ICD-8 code 303 or treatment history with disulfiram).

2.5. Statistical analysis

The study population contributed person-years until either the dateof receiving a diagnosis of hip fracture (ICD-10 code: S72), death, or31 December 2009 whichever came first. The unexposed group waspeople from the general Danish population without a diagnosis ofschizophrenia. Initial comparisons of the exposed and non-exposedpopulations used chi-squared tests for categorical variables andWilcoxon rank sum test for continuous variables.

Incidence rate ratios (IRR) were estimated using Poisson reg-ression analyses. Independent variables in the unmatched analyses

H.J. Sørensen et al.874

were analyzed as categorical variables except for age at inclusion,lifetime consumption (DDD) of antipsychotics and lifetime consumptionof corticosteroids. In the matched analyses among patients with adiagnosis of schizophrenia with and without hip fracture, cases andcontrols were matched 1:3 on age (10 year age groups) and sex(controls were not necessarily unique). Due to this matching weestimated robust standard errors in the Poisson regression analyses.

The effects of long-term exposure were adjusted for number of bed-days, alcohol-related diagnosis, CPDS, antidepressants, anticholinergicsand benzodiazepines. The primary exposures were prescription historyof antipsychotics accumulated between January 01, 1995 and the indexdate. We calculated the average monthly DDDs of PRL-AP and non-PRL-AP (and the sum of these 2 categories). Dose–effect relationships wereexamined for the DDD categories o1; 1–2; 42). All analyses werecarried out in Stata 12 on the Statistics Denmark server via remoteaccess. The chosen level of significance was 5%.

3. Results

3.1. Population based sample

Table 1 shows the demographic characteristics in patients withschizophrenia (exposed) and the unexposed comparison group.The exposed group had significantly higher proportions ofmen, CPDS≥1, and people with short education, early retire-ment pension or with a lifetime diagnosis of alcohol abuse.

Table 1 Distribution of risk factors for hip fracture in patiecontrols (n=3,807,597).

Schizophrenia

N %

SexMale 9.105 59

Age, years (mean, sd) at inclusion 49.9 (13.4)Highest level of education at inclusionPrimary school 8.021 52High school 1.397 9.1Vocational education 2.981 19Short-cycle higher 320 2.1Medium higher 1.005 6.5Long higher 367 2.4NA 1.340 8.7

Any kind of pension at inclusionYes 12.406 80No 2.609 16NA 416 2.7

Somatic co-morbidity 1% with CPDS ≥ 6.900 44

Lifetime exposure, DDD (mean, sd)Any antipsychotic 1.26 1.2Any corticosteroroids 0.01 0.0

Lifetime alcohol abuse 3.542 23

Lifetime other medicationAntidepressants 7.223 46Anticholinergics 9.051 58Benzodiazepines 9.872 64

Age at inclusion was slightly higher and the exposed groupexpectedly received higher DDD's of antipsychotics (slightlyhigher DDD's of corticosteroids), and the percentages ofantidepressants, anticholinergics or benzodiazepines were alsohigher. If hip fracture occurred, the mean age (sd) was 77.3(12.7) years respectively 68.6 (12.8) years in the unexposedand in the schizophrenia group (po0.001).

Table 2 (first column) shows a significantly elevated IRR(1.19 95% CI 1.08–1.13) of hip fracture associated with adiagnosis of schizophrenia after adjustment for all covariatesexcept antipsychotics. After additional adjustment for theeffect of antipsychotics (second column), the effect of schizo-phrenia was not significant (IRR=1.00 95% CI 0.90–1.11).Table 2 (second column) shows independent predictive effectsof antipsychotics (IRR=1.19 95% 1.15–1.24), corticosteroids,antidepressants, anticholinergics, benzodiazepines, CPDS ≥1,lifetime alcohol abuse, education, early retirement, age, whilemale sex showed a protective effect.

3.2. Psychopharmacological treatment ≤4 monthsbefore index date

In the matched sample, Table 3 shows the unadjustedanalyses of major drug classes and individual antipsychotics

nts with schizophrenia (n=15,431) and general population

General population controls

N % p-value

.0 1.639.620 43.1 o0.001

46.9 (17.5)

.0 1.176.031 30.9273.718 7.2

.3 1.166.687 30.6120.161 3.2434.486 11.4168.597 4.4467.917 12.3 o0.001

.4 764.404 20.1

.9 2.910.590 76.4132.603 3.5 o0.001

.7 1.423.110 37.4 o0.001

7 0.06 0.22 o0.0016 0.01 0.07 o0.001

.0 195.308 5.1 o0.001

.8 919.048 24.1 o0.001

.7 31.657 0.8 o0.001

.0 1.018.338 26.7 o0.001

Table 2 Multivariate analysis of predictors of hip fracture among cohort members (n=3,353,771) with available data on allvariables.

First adjustmenta Full adjustmentb

Predictors IRR 95% CI p IRR 95% CI p

Schizophrenia diagnosis 1.19 1.08–1.31 o0.001 1.00 0.90–1.11 0.995Male sex 0.73 0.71–0.74 o0.001 0.73 0.71–0.74 o0.001Age at inclusion 1.08 1.08–1.08 o0.001 1.08 1.08–1.08 o0.001EducationPrimary school Ref – – Ref – –

High school 0.83 0.77–0.88 o0.001 0.83 0.77–0.88 o0.001Vocational education 0.89 0.88–0.91 o0.001 0.89 0.88–0.91 o0.001Short-cycle higher 0.82 0.77–0.88 o0.001 0.82 0.77–0.88 o0.001Medium higher 0.76 0.74–0.79 o0.001 0.77 0.74–0.79 o0.001Long higher 0.73 0.69–0.77 o0.001 0.73 0.69–0.77 o0.001

Early retirement pension 2.34 2.27–2.40 o0.001 2.33 2.26–2.39 o0.001Somatic score ≥1 1.09 1.07–1.11 o0.001 1.09 1.07–1.11 o0.001Lifetime DDD'sAntipsychotic – – – 1.19 1.15–1.24 o0.001Corticosteroroids 1.44 1.35–1.53 o0.001 1.44 1.36–1.53 o0.001

Lifetime alcohol abuse 2.80 2.73–2.88 o0.001 2.80 2.72–2.87 o0.001Lifetime other medicationAntidepressants 1.18 1.16–1.20 o0.001 1.18 1.16–1.20 o0.001Anticholinergics 1.36 1.29–1.43 o0.001 1.29 1.22–1.36 o0.001Benzodiazepines 1.07 1.05–1.08 o0.001 1.06 1.04–1.08 o0.001

aAdjusted for sex, age, education, early retirement pension, CPDS, lifetime DDD's of corticosteroids, lifetime alcohol-relateddiagnosis, antidepressants, anticholinergics, and benzodiazepines.

bAdjusted for sex, age, education, early retirement pension, CPDS, lifetime DDD's of corticosteroids, lifetime alcohol-relateddiagnosis, antidepressants, anticholinergics, and benzodiazepines plus lifetime DDD's of antipsychotics.

875Schizophrenia, antipsychotics and risk of hip fracture

(exposure versus non-exposure) as well as the distribution ofantipsychotics (0, 1, 2, and 42) within 4 months before theindex date of hip fracture. Exposure to anticholinergics,antidepressants, benzodiazepines, antipsychotics (both low-potency, medium potency and high potency), olanzapineand quetiapine were all associated with a significantlyelevated IRR. When treatment with no antipsychotic wasused as the reference, the IRR of 1, 2 and 42 antipsychotic(s) was 1.30 (95% CI 1.10–1.54), 1.68 (95% 1.35–2.07) and1.81 (95% CI 1.28–2.54), respectively.

To further examine whether low potency-, mediumpotency-, high potency- or individual antipsychotics differ-entially predicted the outcome, we performed post hocanalyses (Table 4) based on 827 patients receiving anti-psychotic monotherapy (the largest group receiving olanza-pine was the reference). After adjustment, we did notobserve a differential predictive effect.

3.3. Effects of long-term exposure topsychopharmacological treatment

In the matched sample, we also examined (Table 5, firstcolumn) history of long-term exposure to prolactin-increasing antipsychotics (PRL-AP) and risk of hip fracture.There was a dose-related elevated risk (IRR 1.16 95% CI1.09–1.24). Similarly, we observed a dose-related effect onnon-PRL AP (IRR 1.23 95% CI 1.15–1.32).

In the fully adjusted analyses (third column), both PRL APand non-PRL AP conferred increased risk (IRR 1.12 95% CI1.04–1.21) respectively 1.15 95% CI 1.05–1.25). After adjust-ment, the IRR associated with exposure (in DDD) of anyantipsychotic and fracture risk was 1.13 (95 CI 1.07–1.19).A subdivision of antipsychotics into DDD categories: o1, 1–2,42 confirmed a relationship between higher DDD of anti-psychotics and increased risk of hip fracture (data notshown).

4. Discussion

A diagnosis of schizophrenia was associated with increasedrisk (IRR=1.19 CI 1.08–1.31) in initial analyses adjusted for awide range of other risk factors for hip fracture. However,after further adjustment for the effect of antipsychotictreatment, the effect of schizophrenia was not statisticallysignificant. In addition to a predictive effect of long-termtreatment with antipsychotics (IRR=1.13 95% CI 1.07–1.19),treatment with anticholinergics was associated with asignificantly increased risk of hip fracture, a finding whichadds to previous reports (Kragh et al., 2011; Panula et al.,2009; Vestergaard et al., 2007). Antidepressants were alsoassociated with increased risk of hip fracture which isconsistent with previous findings (Bolton et al., 2008;Howard et al., 2007). Benzodiazepines, were also asso-ciated with hip fracture thus corroborating previousresearch (Vestergaard et al., 2006). Compared with no

Table 3 Analysis (n=2224) of exposure to psychopharmacological medication (within 4 months of the index date) and risk ofhip fracture in patients with schizophrenia.

Medication 4 months prior to fracture N % N % IRRa 95% CI p

Anticholinergics 194 34.9 409 24.5 1.46 1.26–1.70 o0.001Antidepressants 135 24.3 313 18.8 1.29 1.09–1.54 0.004Benzodiazepines 167 30.0 399 23.9 1.34 1.14–1.58 o0.001Antipsychotics 342 61.5 874 52.4 1.42 1.21–1.65 o0.001Low potency 88 15.8 209 12.5 1.29 1.05–1.58 0.013Medium potency 156 28.1 401 24.0 1.19 1.01–1.39 0.036High potency 56 10.1 116 7.0 1.36 1.07–1.73 0.011Clozapine 31 5.6 89 5.3 1.18 0.85–1.63 0.317Aripiprazole 3 0.5 19 1.1 0.60 0.20–1.83 0.370Risperidone 46 8.3 115 6.9 1.18 0.91–1.52 0.222Olanzapine 90 16.2 193 11.6 1.41 1.15–1.73 0.001Quetiapine 24 4.3 42 2.5 1.52 1.11–2.08 0.010Ziprasidone 5 0.9 10 0.6 1.82 0.82–4.06 0.144Sertindole 0 0.0 1 0.1 – – –

Number of antipsychotics0 214 38.5 794 47.6 Ref – –

1 216 38.8 611 36.6 1.30 1.10–1.54 0.0022 99 17.8 208 12.5 1.68 1.35–2.07 o0.00142 27 4.9 55 3.3 1.81 1.28–2.54 0.001

aAdjusted for sex, age at diagnosis, alcohol misuse and somatic score.

Table 4 Analysis (n=827) of exposure to antipsychotics(within 4 months of the index date) and risk of hipfracture in patients with schizophrenia on antipsychoticmonotherapy.

IRR 95% CI p

High dose 0.74 0.41–1.36 0.331Medium dose 0.98 0.70–1.36 0.898Low dose 1.19 0.78–1.80 0.427Clozapine 0.67 0.34–1.32 0.246Risperidone 1.06 0.66–1.72 0.806Olanzapine Ref – –

Quetiapine 0.99 0.51–1.91 0.971Othernn 0.88 0.22–3.54 0.859

n Adjusted for sex, age at diagnosis, alcohol misuse andsomatic score.

H.J. Sørensen et al.876

treatment within 4 months before the index date of hipfracture, treatment with ≥2 antipsychotics was found in17.5% (389/2224) of those with schizophrenia and wasassociated with a significantly higher risk of hip fracture.

4.1. Antipsychotics and hip fracture

Our finding that long-term treatment with prolactin-increasing antipsychotics (PRL-AP) conferred increased riskof hip fracture is consistent with other studies (Howardet al., 2007; Takkouche et al., 2007). Although, Hugenholtzet al. (2005) found a smaller influence of the dose ofprolactin-increasing antipsychotics than was suggested inour study, the findings point in the same direction forprolactin increasing antipsychotics. While Vestergaard

et al. found a small effect of antipsychotics on fracturesas a whole without evidence of a dose–response relationship(Vestergaard et al., 2006), our findings suggest a strongereffect and it is possible that a discrepancy could be due todifferences in methodological approach. To our knowledge,ours is the first study in patients with schizophrenia to findevidence of a dose–response relationship between long-termtreatment with non-prolactin increasing antipsychotics(non-PRL-AP) and risk of hip fracture. These findingswarrant further studies. In elderly patients with variouspsychiatric diagnoses, increased risk of hip fracture wasfound in patients having received long-term treatment witholanzapine as well as in patients treated over longer timewith prolactin raising antipsychotics such as risperidone andhaloperidol (Liperoti et al., 2007). A study of antipsychotictreatment in elderly patients with Parkinsonism and mentalsymptoms observed that the adjusted odds ratio of fracturewas significantly elevated for quetiapine, olanzapine as wellas for the prolactin raising antipsychotic, risperidone (Doreet al., 2009). However, in these studies the effects were notseparated in long term exposure and current exposure in asimilar fashion as done in the present study. Pertaining toTable 3 suggesting that exposure (relative to non-exposure)within 4 months before index date to olanzapine andquetiapine predicts increased risk, these unadjusted esti-mates should be interpreted with caution. Thus, we did notobserve a predictive effect of these substances when werestricted the sample to patients on antipsychotic mono-therapy. Neither in Table 3 nor Table 4, we found anypredictive effect of clozapine, a potent alpha-1 noradre-nergic and histamine antagonist, but it was not at this pointpossible to examine if this could be a result of psychiatristsnot prescribing clozapine to high-risk patients (Nielsenet al., 2010a, 2012). It is possible that the findings in

877Schizophrenia, antipsychotics and risk of hip fracture

Table 3 might be due to some patients with schizophreniaare being exposed to antipsychotic polypharmacy andpossibly a higher total dose of antipsychotics than peopleon antipsychotic monotherapy. Psychiatrists might havebeen more inclined to prescribe higher doses of olanzapineor quetiapine as part of a clinical decision to beginantipsychotic polytherapy. Some of these patients mayfurthermore have been exposed to another more generaltype of antipsychotic polypharmacy where adjunctive med-ications are prescribed for co-morbid conditions (Baandrupet al., 2010).

4.2. Strengths and limitations

The present nationwide register linkage case control studyis among the largest conducted in patients diagnosed withschizophrenia and population controls investigating if thedisorder is associated with increased risk for hip fracture.The access to the prescription database from 1995 allowed aminimum of 5 years of pharmacy history. The study isnaturalistic in nature and design, i.e. our data on theassociation between antipsychotics and subsequent risk ofdeveloping hip fracture reflects the association in clinicalpractice in which doctors choose to prescribe antipsycho-tics, continue or switch to another drug, and monitor peoplebased on their individual current knowledge. Psychiatricand medical care in Denmark is well developed so thatpeople can easily come into contact with general practi-tioners or specialists in psychiatry and bias by socioeco-nomic differences are unlikely. Data from the prescriptiondatabase is close to 100% accurate. Nationwide access toregistry data from the relevant registries ensured that allincident cases of hip fracture were included. There are alsoseveral limitations, including the observational study designwith no possibility of determining causality, lack of rando-mized treatment assignment, and potential influence ofunmeasured and unknown confounders, that cannot becontrolled for, such as the severity of schizophrenia,inactivity, body mass index, diet and smoking. It was notpossible to adjust for non-compliance or partial complianceor to take all treatment discontinuations in the period up toindex date into account. DDD was used to estimate theindividual effect of each antipsychotic drug but especiallyfor the newer antipsychotic drugs the assigned dose hasbeen questionable (Nielsen et al., 2010b). Medicationpurchased before January 01, 1995 could not be taken intoaccount.

4.3. Clinical implications

The results of our study suggest a potential for preventivemeasures for hip fracture in schizophrenia. Attention couldfor instance be drawn to patients with an early debut ofschizophrenia and those who receive complex psychophar-macological treatment. More emphasis on rational psycho-pharmacological treatment and on the cost-benefits ofpolypharmacy in treatment resistant schizophrenia isanother aspect that could potentially influence the absoluterisk of hip fracture in patients with schizophrenia. In thiscontext, high-risk patients may also be the somatic frailpatient: some of who might have latent gait and balance

deficits (Schiffman et al., 2009; Sullivan et al., 2004). Othermeasures such as screening of high-risk groups for earlysigns of parkinsonism, balance deficits, osteoporosis as wellas perhaps in some cases earlier introduction of prophylac-tic medical treatment. Interventions aimed at changes inlifestyle might also have an impact on the risk of hipfracture in patients treated with antipsychotics and inpatients with schizophrenia.

Role of funding source

No funding source.

Contributors

J. Nielsen designed the study. S.O.W. Jensen carried out theanalysis and H. Sorensen wrote the first draft of the manuscript.All authors edited the manuscript.

Conflict of interest

J. Nielsen has received research grants from H. Lundbeck, Pfizerand Chempaq for clinical trials and received speaking fees fromBristol-Myers Squibb, Astra Zeneca, Lundbeck, Janssen–Cilag,Hemocue, and Eli-Lilly.

Other authors report no conflict of interest.

Acknowledgment

None.

References

Abraham, G., Halbreich, U., Friedman, R.H., Josiassen, R.C., 2003.Bone mineral density and prolactin associations in patients withchronic schizophrenia. Schizophr. Res. 59, 17–18.

Baandrup, L., Gasse, C., Jensen, V.D., Glenthoj, B.Y., Nordentoft,M., Lublin, H., Fink-Jensen, A., Lindhardt, A., Mortensen, P.B.,2010. Antipsychotic polypharmacy and risk of death from naturalcauses in patients with schizophrenia: a population-basednested case-control study. J. Clin. Psychiatry 71, 103–108.

Bolton, J.M., Metge, C., Lix, L., Prior, H., Sareen, J., Leslie, W.D.,2008. Fracture risk from psychotropic medications: apopulation-based analysis. J. Clin. Psychopharmacol. 28,384–391.

Cumming, R.G., Nevitt, M.C., Cummings, S.R., 1997. Epidemiologyof hip fractures. Epidemiol. Rev. 19, 244–257.

Cummings, S.R., Melton, L.J., 2002. Epidemiology and outcomes ofosteoporotic fractures. Lancet 359, 1761–1767.

Dore, D.D., Trivedi, A.N., Mor, V., Friedman, J.H., Lapane, K.L.,2009. Atypical antipsychotic use and risk of fracture in personswith Parkinsonism. Movement Disord. 24, 1941–1948.

Halbreich, U., Palter, S., 1996. Accelerated osteoporosis in psy-chiatric patients: possible pathophysiological processes. Schi-zophr. Bull. 22, 447–454.

Howard, L., Kirkwood, G., Leese, M., 2007. Risk of hip fracture inpatients with a history of schizophrenia. Br. J. Psychiatry 190,129–134.

Hugenholtz, G.W., Heerdink, E.R., van Staa, T.P., Nolen, W.A.,Egberts, A.C., 2005. Risk of hip/femur fractures in patients usingantipsychotics. Bone 37, 864–870.

H.J. Sørensen et al.878

Kragh, A., Elmstahl, S., Atroshi, I., 2011. Older adults' medicationuse 6 months before and after hip fracture: a population-basedcohort study. J. Am. Geriatr. Soc. 59, 863–868.

Liperoti, R., Onder, G., Lapane, K.L., Mor, V., Friedman, J.H.,Bernabei, R., Gambassi, G., 2007. Conventional or atypicalantipsychotics and the risk of femur fracture among elderlypatients: results of a case-control study. J. Clin. Psychiatry 68,929–934.

Mors, O., Perto, G.P., Mortensen, P.B., 2011. The Danish PsychiatricCentral Research Register. Scand. J. Public Health 39, 54–57.

Munk-Jorgensen, P., Mortensen, P.B., 1997. The Danish PsychiatricCentral Register. Dan. Med. Bull. 44, 82–84.

Nielsen, J., Dahm, M., Lublin, H., Taylor, D., 2010a. Psychiatrists'attitude towards and knowledge of clozapine treatment. J.Psychopharmacol. 24, 965–971.

Nielsen, J., le Quach, P., Emborg, C., Foldager, L., Correll, C.U.,2010b. 10-Year trends in the treatment and outcomes ofpatients with first-episode schizophrenia. Acta Psychiatr. Scand.122, 356–366.

Nielsen, J., Foldager, L., Meyer, J.M., 2009. Increased use ofantibiotics in patients treated with clozapine. Eur. Neuropsy-chopharmacol. 19, 483–486.

Nielsen, J., Meyer, J.M., 2012. Risk factors for ileus in patients withschizophrenia. Schizophr. Bull. 38, 592–598.

Nielsen, J., Roge, R., Schjerning, O., Sorensen, H.J., Taylor, D.,2012. Geographical and temporal variations in clozapine pre-scription for schizophrenia. Eur. Neuropsychopharmacol..

Ostergaard, S.D., Petrides, G., Dinesen, P.T., Skadhede, S., Bech, P.,Munk-Jorgensen, P., Nielsen, J., 2013. The association betweenphysical morbidity and subtypes of severe depression. Psychother.Psychosom. 82, 45–52.

Panula, J., Puustinen, J., Jaatinen, P., Vahlberg, T., Aarnio, P.,Kivela, S.L., 2009. Effects of potent anticholinergics, sedatives

and antipsychotics on postoperative mortality in elderly patientswith hip fracture: a retrospective, population-based study.Drugs Aging 26, 963–971.

Partti, K., Heliovaara, M., Impivaara, O., Perala, J., Saarni, S.I.,Lonnqvist, J., Suvisaari, J.M., 2010. Skeletal status in psychoticdisorders: a population-based study. Psychosom. Med. 72,933–940.

Pouwels, S., van Staa, T.P., Egberts, A.C., Leufkens, H.G., Cooper, C.,de Vries, F., 2009. Antipsychotic use and the risk of hip/femurfracture: a population-based case-control study. Osteoporosis Int..

Schiffman, J., Sorensen, H.J., Maeda, J., Mortensen, E.L., Victor-off, J., Hayashi, K., Michelsen, N.M., Ekstrom, M., Mednick, S.,2009. Childhood motor coordination and adult schizophreniaspectrum disorders. Am. J. Psychiatry 166, 1041–1047.

Sullivan, E.V., Rosenbloom, M.J., Pfefferbaum, A., 2004. Balance andgait deficits in schizophrenia compounded by the comorbidity ofalcoholism. Am. J. Psychiatry 161, 751–755.

Takkouche, B., Montes-Martinez, A., Gill, S.S., Etminan, M., 2007.Psychotropic medications and the risk of fracture: a meta-analysis. Drug Saf. 30, 171–184.

Uggerby, P., Nielsen, R.E., Correll, C.U., Nielsen, J., 2011. Char-acteristics and predictors of long-term institutionalization inpatients with schizophrenia. Schizophr. Res. 131, 120–126.

Uggerby, P., Ostergaard, S.D., Roge, R., Correll, C.U., Nielsen, J.,2013. The validity of the schizophrenia diagnosis in the DanishPsychiatric Central Research Register is good. Dan. Med. J. 60,A4578.

Vestergaard, P., Rejnmark, L., Mosekilde, L., 2006. Anxiolytics,sedatives, antidepressants, neuroleptics and the risk of frac-ture. Osteoporosis Int. 17, 807–816.

Vestergaard, P., Rejnmark, L., Mosekilde, L., 2007. Fracture riskassociated with parkinsonism and anti-Parkinson drugs. Calcif.Tissue Int. 81, 153–161.