The United Kingdom General Practice Research Database (GPRD) is an office-based, computer-generated, medical resource designed from its inception to be used for epidemiologic research. A distinct version of the GPRD is maintained by the Boston Collaborative Drug Surveillance Program and has been the source of more than 130 scientific articles primarily addressing drug safety issues. We reviewed evidence related to the validity of the GPRD. Specifically, with our extensive experience with this automated database, we evaluated the quality and completeness of the data that it contains.

Problems associated with pharmacotherapy (in particular, medication errors and adverse drug events) are frequent and are associated with increased costs for treatment. Analysis of original publications published between 1990 and 2005 on the topics of medication errors and/or adverse drug events in hospitalised patients, focusing on the frequency of, risk factors for and avoidance of such problems associated with pharmacotherapy, indicated that medication errors occurred in a mean of 5.7% of all episodes of drug administration, but with a high variability among the 35 studies retrieved. This variability was explained by the methods by which medication errors were detected (systematic screening of patients versus chart review or spontaneous reporting) and by the way drugs were administered (intravenously administered drugs are associated with the highest error frequencies). Errors occurred throughout the whole medication process, with administration errors accounting for more than half of all errors. Important risk factors included insufficient pharmacological knowledge of health professionals, errors in the patient charts or documentation by nurses and inadequate pharmacy services.Adverse events or reactions, on the other hand, affected 6.1 patients per 100 hospitalised and also showed a high variability among the 46 studies retrieved. This variability could also be explained by the different methods of assessment of the frequency of adverse drug events or reactions, as well as by the different wards on which the studies were performed. Important risk factors for adverse drug events or reactions included polypharmacy, female sex, drugs with a narrow therapeutic range, renal elimination of drugs, age >65 years and use of anticoagulants or diuretics. Since medication errors are strong risk factors for preventable adverse drug events or reactions, strategies have to be put in place for their reduction. Such strategies include ensuring that all persons involved in the medication process (nurses, pharmacists and physicians) have good pharmacological knowledge, computerisation of the entire medication process, and the engagement of a sufficient number of clinical pharmacists on the wards.

Context Animal studies suggest that the ␤-blocker propranolol increases bone formation, but data on whether use of ␤-blockers (with or without concomitant use of thiazide diuretics) is associated with reduced fracture risk in humans are limited. Objective To determine whether use of ␤-blockers alone or in combination with thiazides is associated with a decreased risk of fracture in adults. Design, Setting, and Participants Case-control analysis using the UK General Practice Research Database (GPRD). The study included 30601 case patients aged 30 to 79 years with an incident fracture diagnosis between 1993 and 1999 and 120819 controls, matched to cases on age, sex, calendar time, and general practice attended. Main Outcome Measures Odds ratios (ORs) of having a fracture in association with use of ␤-blockers or a combination of ␤-blockers with thiazides. Results The most frequent fractures were of the hand/lower arm (n=12837 [42.0%]) and of the foot (n=4627 [15.1%]). Compared with patients who did not use either ␤-blockers or thiazide diuretics, the OR for current use of ␤-blockers only (Ն3 prescriptions) was 0.77 (95% confidence interval [CI], 0.72-0.83); for current use of thiazides only (Ն3 prescriptions), 0.80 (95% CI, 0.74-0.86); and for combined current use of ␤-blockers and thiazides, 0.71 (95% CI, 0.64-0.79). Data were adjusted for smoking; body mass index; number of practice visits; and use of calcium channel blockers, angiotensin-converting enzyme inhibitors, antipsychotics, antidepressants, statins, antiepileptics, benzodiazepines, corticosteroids, and estrogens. Conclusions Our data suggest that current use of ␤-blockers is associated with a reduced risk of fractures, taken alone as well as in combination with thiazide diuretics. Many elderly patients with hypertension who are at risk of developing osteoporosis may potentially benefit from combined therapy with ␤-blockers and thiazides.

This study suggests that current exposure to statins is associated with a decreased risk of bone fractures in individuals age 50 years and older. This finding has a potentially important public health impact and should be confirmed further in controlled prospective trials. JAMA. 2000;283:3205-3210

Unfortunately, there is no endogenous marker for hepatic clearance that can be used as a guide for drug dosing. In order to predict the kinetic behaviour of drugs in cirrhotic patients, agents can be grouped according to their extent of hepatic extraction. For drugs with a high hepatic extraction (low bioavailability in healthy subjects), bioavailability increases and hepatic clearance decreases in cirrhotic patients. If such drugs are administered orally to cirrhotic patients, their initial dose has to be reduced according to hepatic extraction. Furthermore, their maintenance dose has to be adapted irrespective of the route of administration, if possible, according to kinetic studies in cirrhotic patients. For drugs with a low hepatic extraction, bioavailability is not affected by liver disease, but hepatic clearance may be affected. For such drugs, only the maintenance dose has to be reduced, according to the estimated decrease in hepatic drug metabolism. For drugs with an intermediate hepatic extraction, initial oral doses should be chosen in the low range of normal in cirrhotic patients and maintenance doses should be reduced as for high extraction drugs. In cholestatic patients, the clearance of drugs with predominant biliary elimination may be impaired. Guidelines for dose reduction in cholestasis exist for many antineoplastic drugs, but are mostly lacking for other drugs with biliary elimination. Dose adaptation of such drugs in cholestatic patients is, therefore, difficult and has to be performed according to pharmacological effect and/or toxicity. Importantly, the dose of drugs with predominant renal elimination may also have to be adapted in patients with liver disease. Cirrhotic patients often have impaired renal function, despite a normal serum creatinine level. In cirrhotic patients, creatinine clearance should, therefore, be measured or estimated to gain a guideline for the dosing of drugs with predominant renal elimination. Since the creatinine clearance tends to overestimate glomerular filtration in cirrhotic patients, the dose of a given drug may still be too high after adaptation to creatinine clearance. Therefore, the clinical monitoring of pharmacological effects and toxicity of such drugs is important. Besides the mentioned kinetic changes, the dynamics of some drugs is also altered in cirrhotic patients. Examples include opiates, benzodiazepines, NSAIDs and diuretics. Such drugs may exhibit unusual adverse effects that clinicians should be aware of for their safe use. However, it is important to realise that the recommendations for dose adaptation remain general and cannot replace accurate clinical monitoring of patients with liver disease treated with critical drugs.

Using a computerised drug-interaction program, a high proportion of patients was detected with at least one potential DDI in the medication prescribed at discharge. However, the proportion of DDIs associated with potentially relevant clinical consequences appeared to be relatively low.

Despite applying uniform methods, variations in the prevalence of antidepressant prescribing were obvious in the different populations. Database characteristics and clinical factors may both explain these variations.

No abstract