CNS Drugs 2003; 17 (11): 825-837

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 825 1. Psychotropic Drugs as a Risk Factor for Hip Fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 826 2. Methodological Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 827 2.1 Hospital-Based Case-Control Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 827 2.2 Confounding by Indication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 828 2.3 Measurement of Benzodiazepine Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 828 3. Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 829 3.1 Use of Benzodiazepines in the Community or Nursing Homes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 830 3.2 Benzodiazepine Half-Life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 830 3.3 Dose of Benzodiazepines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 831 3.4 Oxidative and Nonoxidative Benzodiazepines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 832 3.5 Recently Initiated and Long-Term Use of Benzodiazepines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 832 4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 832 4.1 Pharmacology of Benzodiazepines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 833 5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834 A hip fracture epidemic is occurring in developed countries in association with Abstract population aging. The increasing number of people with a hip fracture has major implications for clinicians and health service managers. More importantly, a hip fracture is a devastating event in the life of an older person, as it often leads to loss of independence and death. Identification of risk factors for hip fracture is an essential first step towards prevention. The use of psychotropic medications is an established risk factor for hip fracture. The purpose of this article is to systematically review epidemiological studies of the relationship between use of benzodiazepines and risk of hip fracture and, then, to see how the findings of these studies fit with what is known about the pharmacology of benzodiazepines. 826 Cumming & Le Couteur Eleven primary epidemiological studies were identified. The results of these studies were not consistent; however, the inconsistency appeared to be almost entirely explained by research design. The studies that did not show an association between increased hip fracture risk and benzodiazepine use were nearly all hospital-based case-control studies, a type of study that often lacks validity because of the difficulty of finding an appropriate control group. After excluding the hospital-based case-control studies, all but one of the remaining seven studies found that use of benzodiazepines was associated with an increased risk of hip fracture that varied between 50% and 110%. The only higher quality study that did not find an association between benzodiazepine use and hip fracture was also the only study conducted entirely in nursing homes. There was no evidence that the risk of hip fracture differed between shortand long-acting benzodiazepines. People using higher doses of benzodiazepines and those who had recently started using benzodiazepines were at the highest risk of hip fracture. In very old people, there was some preliminary evidence that benzodiazepines that undergo oxidation in the liver may be associated with a higher risk of hip fracture than other benzodiazepines. The epidemiological evidence strongly suggests that the use of benzodiazepines by older people increases their risk of hip fracture by at least 50%. The benefits of benzodiazepines for older people are unclear. Given the high morbidity and mortality of hip fracture, it can be concluded that older people should rarely be prescribed benzodiazepines and that many older people already taking these drugs should have them withdrawn under appropriate supervision. Population aging is associated with a hip fracture their prefracture levels of physical function and social activities.[2] epidemic in developed countries. For example, in Australia in 1996 there were 15 000 hip fractures. 1. Psychotropic Drugs as a Risk Factor As the percentage of the population that is elderly for Hip Fracture increases, it is estimated that the number of hip fractures will rise to 26 000 by 2016, 34 000 by There is clearly an urgent need to identify risk factors for hip fracture so that effective preventive 2026 and 60 000 by 2051.[1] The rapidly increasing strategies can be implemented. The use of psychonumber of people with a hip fracture has obvious tropic medications is one risk factor for hip fracture implications for health service providers and manthat has received a great deal of research attention. agers. More importantly, hip fracture is a devastatThere is now good evidence that all classes of ing event in the life of an older person. The 1-year psychotropics are associated with an increased risk mortality rate after hip fracture is about 20% and, of of hip fracture. Barbiturates were implicated as a those living in the community at the time of their hip cause of hip fracture as long ago as 1977[3] and, 10 fracture, 20% will be admitted to a nursing home years later, Ray and colleagues[4] showed that the following the hip fracture.[2] Of those returning to use of tricyclic antidepressants and antipsychotics live in the community, the majority will never regain each doubled the risk of hip fracture. More recently,  Adis Data Information BV 2003. All rights reserved. CNS Drugs 2003; 17 (11) Benzodiazepines and Hip Fractures 827 the SSRI antidepressants have been found to have a A computerised Medline search was conducted similar effect.[5] for the period January 1966 to September 2002. Search terms used were ‘hip fractures’ and ‘benzoAmong the psychotropic medications, benzodiazepines’ or ‘psychotropic drugs’. This search diazepines have received particular attention beproduced 65 articles. We also searched our own cause of their extensive use in older people. In collection of papers on hip fracture epidemiology. recent surveys, the prevalence of benzodiazepine Articles reporting results for the broad class of sedause has ranged from 17% in Australia[6] and Iretives and/or hypnotics, without providing specific land[7] to 23% in Canada[8] to 32% in France.[9] results for benzodiazepines, are not considered in Long-term benzodiazepine use appears to become detail here. more common with increasing age. For example, the To assess the possible effects of study design on Canadian survey found that 20% of those aged study findings, studies were classified into the three 60–69 years had used benzodiazepines long term main observational epidemiological study types: cocompared with 30% of those ≥80 years.[8,10] Older hort studies, population-based case-control studies people are also frequently administered benzodiazeand hospital-based case-control studies. An article pines for sedation during hospital admissions. In one reporting changes in hip fracture rates after the study of older patients in medical wards, 20% were benzodiazepine prescribing policy was changed in prescribed benzodiazepines, mostly temazepam; New York state, USA, was excluded because of the over half of these patients had not been taking weakness of this type of pre-/poststudy design for benzodiazepines prior to hospital admission.[11] assessing causal relationships.[14] There is now good evidence that benzodiazepines cause falls, which are commonly the cause of hip 2. Methodological Issues fracture. A meta-analysis of observational epidemiological studies found that use of benzodiazepines 2.1 Hospital-Based Case-Control Studies was associated with a 50% increased risk of falling.[12] Furthermore, a recently published randomThe potential for bias in epidemiological studies ised trial found that reducing the dose of psychois strongly dependent on the study type. Hospitaltropic medications (benzodiazepines or antidepresbased case-control studies are generally considered sants) reduced the risk of falling by 66%.[13] the weakest type of epidemiological study because This article provides a detailed review of epideof the difficulty of choosing appropriate control miological studies of the relationship between the subjects. In hospital-based case-control studies, both use of benzodiazepines and the risk of hip fracture. the cases (here, people who have had a hip fracture) We have addressed a range of issues including the and controls (people who have not had a hip fracrelative safety of shortversus long-acting benzoture) are recruited from among hospitalised patients. diazepines, the effects of lower versus higher doses While it is appropriate for patients with a hip fracof benzodiazepines, recent and long-term use, and ture to be recruited from hospitals (because nearly benzodiazepines that undergo phase I oxidative metall people who have a hip fracture are admitted to abolism in the liver versus those that undergo phase hospital), recruiting control individuals from hospiII conjugative metabolism. Some discussion of the tals is problematic. The purpose of the control group literature addressing benzodiazepine use and falls is is to provide data on the frequency of exposure in included, but no attempt has been made to be comthe population from which the hip fracture patients prehensive. came, to assess whether their exposure is higher (or  Adis Data Information BV 2003. All rights reserved. CNS Drugs 2003; 17 (11) 828 Cumming & Le Couteur lower) than usual. Specifically, the controls need to and, perhaps, psychosocial circumstances when prebe selected in a way that ensures that their use of scribing. This is known as confounding by indicabenzodiazepines is representative of benzodiazepine tion. use among older people living in the population of In observational epidemiological studies of interest, usually the general population. It is clear benzodiazepines and hip fractures, confounding by that patients in hospital, who are unwell, are not indication is most likely to occur when comparing representative of the general population, which different types and different doses of benzodiazecomprises a mix of healthy and unhealthy people. If pines. For people at the highest risk for hip fracture, use of benzodiazepines is more common among the doctors might choose to prescribe particular types sort of people admitted to hospital (for health proband doses of benzodiazepines that they perceive as lems other than hip fracture), then hospital-based safer, and vice versa. The net effect would be to case-control studies of the relationship between make lower risk drugs appear more harmful (bebenzodiazepine use and hip fractures will tend to cause they tend to be used by people at high risk) underestimate, or even miss entirely, any associaand higher risk drugs appear safer (because they tion. On the other hand, if the control group included tend to be used by people at lower risk). Doctors a disproportionate number of people with a disease might also decide not to prescribe benzodiazepines in which benzodiazepine use is contraindicated, at all for people at the highest risk of falling, leading such as people with severe respiratory disease, then to an underestimation in observational epidemiologa hospital-based case-control study might exaggerical studies of the strength of the association beate the size of any benzodiazepine-hip fracture assotween benzodiazepines and hip fractures. ciation. The selection bias of confounding by indication The potential for bias in hospital-based casecan only be avoided completely by conducting rancontrol studies means that their findings must be domised trials. There have been no randomised treated with great caution. (Note that the recruitment trials of benzodiazepines and hip fractures, and it is of controls from hospitalised patients is entirely unlikely that there ever will be. In the absence of appropriate in case-control studies investigating randomised trials, investigators can try to control for causes of hip fractures that occur in hospital, as the confounding by indication by measuring a range of control group provides data on the usual use of risk factors and then adjusting for these potential benzodiazepines by patients while in hospital.) confounders using appropriate statistical methods. Most of the studies included in this review have 2.2 Confounding by Indication controlled for at least some risk factors for hip fracture when assessing the relationship between A major challenge in conducting observational benzodiazepine use and hip fracture. epidemiological studies is the fact that subjects selfselect themselves to become exposed or not exposed 2.3 Measurement of Benzodiazepine Use to the factor of interest, which might be a lifestylerelated behaviour (such as smoking, physical acInaccurate exposure measurement in epidemiotivity or diet) or a medication (such as estrogen logical studies tends to cause bias towards the null replacement therapy or benzodiazepines). This sehypothesis (assuming the inaccuracy is equally likelection bias is a particular problem in studies of ly in patients and controls). Assessment of exposure medications, where the decision to take a drug into benzodiazepines in epidemiological studies is volves a doctor, who considers the patient’s medical usually based on self-report or medical records. To  Adis Data Information BV 2003. All rights reserved. CNS Drugs 2003; 17 (11) Benzodiazepines and Hip Fractures 829 determine the accuracy of these methods, Moore et 3. Findings al.[15] compared the results of medical records, selfEleven primary studies that reported the risk of report and plasma concentrations of benzodiazehip fracture in subjects who used benzodiazepines pines in 1136 older people admitted to a French compared with those who did not were identified.[16-26] Study design characteristics and results of hospital. With the plasma concentration results as these 11 studies are shown in table I. There were the reference standard, self-report had a sensitivity four hospital-based case-control studies, six populaof 80% and a specificity of 83%, while medical tion-based case-control studies and one cohort records had a sensitivity of 71% and a specificity of study. One of the population-based case-control studies was concerned only with hip fractures occur89%. This degree of misclassification means that ring in hospital (and so should more correctly be epidemiological studies will tend to underestimate called a case-control study with a well defined study the strength of the association between use of benzobase).[22] Four of the other population-based casediazepines and risk of hip fracture. control studies used a record linkage methodology, Table I. Epidemiological studies of the relationship between benzodiazepine use and risk of hip fracture Reference Year Country and time period Sample size Resultsa