A growing number of recent reports have suggested that there is an association between decreased bone mineral density and schizophrenia (Reference Halbreich, Rojansky and PalterHalbreich et al, 1995; Reference Abraham, Halbreich and FriedmanAbraham et al, 2003), particularly in patients treated with psychotropic medication (Reference Halbreich and PalterHalbreich & Palter, 1996; Reference O'Keane and MeaneyO'Keane & Meaney, 2005). However, these studies have not examined whether this is clinically significant, leading to an increased risk of fractures. The prevalence of functional psychoses is 4 per 1000 (Reference Jenkins, Lewis and BebbingtonJenkins et al, 1997), with approximately half of that figure relating to schizophrenia. The identification of schizophrenia as a risk factor for osteoporosis would therefore have important public health implications.

Hip fracture is the most important fracture in terms of patient morbidity and mortality and for utilisation of health service resources (Reference Cumming, Nevitt and CummingsCumming et al, 1997). As fracture is common in the older population, a small increase in the risk of fracture associated with psychotic disorders could have a considerable public health effect. We therefore chose to investigate whether there was an association between schizophrenia and hip fractures, using a case–control study design with data from a UK primary care data-set, the General Practice Research Database (GPRD). Our hypothesis was that patients with a history of schizophrenia would have a significantly increased risk of antipsychotic-induced osteoporotic hip fractures compared with a control group matched for age and general practice.

METHOD

RESULTS

There were 16 341 cases of hip fracture and 29 889 controls. There were two controls per case for 13 548 of the cases and one control per case for the remaining 2793. The mean age of the cases and controls at the time of first fracture was 79 years (s.d.=12), and 12 854 (79%) of the cases were female. Variables identified as having an association with hip fracture in the univariate analysis included cerebrovascular disease (OR=1.89, 95% CI 1.76–2.03), blood disorders (OR=1.85, 95% CI 1.73–1.97), intestinal disorders (OR=1.7, 95% CI 1.69–1.86), eye disorders (OR=1.11, 95% CI 1.06–1.17), ear disorders (OR=1.02, 95% CI 0.97–1.08), urinary disorders (OR=1.50, 95% CI 1.42–1.57), laxatives (OR=1.97, 95% CI 1.89–2.07), opioid analgesics (OR=1.94, 95% CI 1.81–2.07), non-opioid analgesics (OR=1.85, 95% CI 1.77–1.93), inhaled corticosteroids (OR= 1.26, 95% CI 1.17–1.36) and injected corticocosteroids (OR=1.07, 95% CI 0.96–1.19). More details of variables of particular interest to this study, including psychotropic medications, are presented in Table 1. Medication with heparin was considered for inclusion in the analysis, but this was ruled out as only 8 cases and 5 controls were exposed to this drug. Fitting bivariate models for each of these variables along with the schizophrenia variable identified only one having a substantial effect on the odds ratio attached to schizophrenia diagnosis, namely having ever had a prescription for a neuroleptic. There was a significant association between having had more than one fall and hip fracture (OR=3.79, 95% CI 3.43–4.19). However, this was considered to be potentially on the causal pathway between schizophrenia diagnosis (and subsequent neuroleptic medication prescription) and hip fracture, assuming such an association exists, and consequently was not included in further analysis.

Although schizophrenia was significantly associated with hip fractures in the univariate analysis, this association was not significant when antipsychotic medication was added to a multivariate model as a potential confounder. A significant interaction was found between gender and having ever had a prescription for neuroleptic medication (χ²=4.15, d.f.=1, P=0.042, likelihood ratio test). No such interaction was found for schizophrenia diagnosis. The model was therefore fitted separately for women (Table 2) and men (Table 3). There were some differences between the two gender-based models. Only in the female model were diseases of the urinary system, having ever had inhaled corticosteroids and diseases of the ear and mastoid process found to be significant. Only in the male model were alcohol consumption above recommended levels and diseases of the intestines and peritoneum found to be significant.

Dementia had been diagnosed in 3461 patients of this data-set – 1815 (11%) cases and 1646 (6%) controls; χ²=478.36, d.f.=1, P<0.01. Antipsychotic medication could have been prescribed for behavioural disturbance in dementia, so patients receiving such medication were excluded and the analyses repeated: this made negligible difference to any result (further details available from the authors upon request). The analyses were repeated using the ‘cluster’ option in Stata, but there was no evidence of clustering of fractures by general practice (further information available from the authors upon request).

There was also a significant interaction between age at the time of the first fracture (categorised) and having ever had a prescription for neuroleptic medication (χ²=28.27, d.f.=3, P<0.001, likelihood ratio test). The interaction was most marked for those aged under 85 years (n=31 071; OR for neuroleptic medication 2.29, 95% CI 2.06–2.53) compared with those aged 85 years and over (n=15 113; OR for neuroleptic medication 1.66, 95% CI 1.49–1.85). Since there was an interaction for both age and gender, the model was also fitted separately with the data-set stratified by both these variables into women under 85 years old (n=23 289; OR for neuroleptic medication 2.31, 95% CI 2.06–2.58); women 85 years or older (n=13 027; OR=1.65, 95% CI 1.47–1.86); men under 85 years old (n=7788; OR=2.42, 95% CI 1.89–3.09); and men 85 years old or older (n=2098; OR=2.12, 95% CI 1.51–2.99). Odds ratios were calculated for each of the individual neuroleptic medications which went to make up the combined neuroleptic variable, and are presented in Table 4. Also included here are figures for prochlorperazine, which was not included as part of the combined neuroleptic variable as it was felt that this might have been used largely as an anti-emetic. Atypical antipsychotic medications (which were only prescribed in the last few years of this data-set), were also considered separately, as there were too few prescribed to include in a combined antipsychotic variable, and odds ratios for these are also presented.

DISCUSSION

We found an independent significant association between hip fracture and prolactin-raising antipsychotic medications used prior to the index fracture. The relationship between schizophrenia and hip fracture was confounded by neuroleptics, and the relationship therefore appears to be owing to the effect of neuroleptics rather than diagnosis per se. There is increasing evidence that long-term antipsychotic-induced hyperprolactinaemia is associated with bone mineral density loss which appears to be mediated by secondary hypogonadism (Reference Bilici, Cakirbay and GulerBilici et al, 2002; Reference O'Keane and MeaneyO'Keane & Meaney, 2005). However, ours is the first study, to our knowledge, to demonstrate an association between antipsychotic medication and a clinically relevant outcome of osteoporosis, i.e. fracture of the neck of femur or hip fracture. A history of neuroleptic use was found to be the most significant predictor of an increased risk of hip fracture, with the exception of being underweight.

A number of recent studies have indicated that low bone mineral density is highly prevalent in people with a chronic psychiatric disorder treated with prolactin-raising antipsychotic medication (Reference Bilici, Cakirbay and GulerBilici et al, 2002; Reference Liu-Seifert, Kinon and AhlLiu-Seifert et al, 2004). Meaney et al (Reference Meaney, Smith and Howes2004) also reported that higher doses of potent typical and atypical antipsychotic medications that block dopamine D₂ receptors were associated with increased rates of both hyperprolactinaemia and bone mineral density loss. However, possibly because of a lack of research into the secondary consequences of antipsychotic-induced hyperprolactinaemia, the World Health Organization does not include antipsychotic drugs in its list of prescribed drugs associated with the development of osteoporosis.

Although postmenopausal women are generally most at risk of osteoporosis, possibly owing to low serum oestradiol concentrations (Reference Cummings, Browner and BauerCummings et al, 1998), we found a higher risk of fracture associated with neuroleptic medication in men. This result is in agreement with several studies of psychiatric patients, which found significantly lower bone mineral density in men than women associated with neuroleptic use (Reference Halbreich, Rojansky and PalterHalbreich et al, 1995; Reference Hummer, Malik and GasserHummer et al, 2005). These gender differences may be owing to the age differences in onset of schizophrenia (Reference Hafner, an der Heiden and BehrensHafner et al, 1998): men have an age at onset approximately 5 years younger than that in women, and illness-related factors including medication will therefore have had a longer-lasting impact on male patients. An alternative explanation suggested by Hummer & Huber (Reference Hummer and Huber2004) is that women with schizophrenia take better care of themselves with regard to adequate nutrition and exercise than men and therefore have less osteoporosis. Unfortunately the data are not available here to test either hypothesis.

There are a number of other known risk factors for osteoporosis among patients with schizophrenia which may be acting as confounders here, including inadequate exercise and exposure to sunshine, poor nutrition, cigarette smoking and polydipsia (Reference Naidoo, Goff and KlibanskiNaidoo et al, 2003). Of these, only smoking could be controlled for in this analysis and was found to be a significant factor, but the association was not as strong as with neuroleptic medication. Other mechanisms may also be relevant in causing hip fracture: for example, neuroleptic medications are known to cause sedation, orthostatic hypotension and extrapyramidal side-effects, which may predispose some patients on these treatments to falls (Reference Misra, Papakostas and KlibanskiMisra et al, 2004).

We used the GPRD, a large UK primary care data-set, which provided one of the largest data-sets of hip fracture. Like other studies, we found an increased risk of hip fracture to be associated with smoking (Reference Cumming, Nevitt and CummingsCumming et al, 1997), low body mass index (Reference Farahmand, Michaelsson and BaronFarahmand et al, 2000), alcohol intake (for men) (Reference Yuan, Dawson and CooperYuan et al, 2001) and anticonvulsants (Reference Kinjo, Setoguchi and SchneeweissKinjo et al, 2005), and obesity to be protective against hip fracture (Reference Farahmand, Michaelsson and BaronFarahmand et al, 2000), giving a high level of face validity to this study.