Article Text


Systematic review
Use of sedatives and hypnotics, antidepressants and benzodiazepines in older people significantly increases their risk of falls
  1. Sirpa Hartikainen1,
  2. Eija Lönnroos2
  1. 1Kuopio Research Centre of Geriatric Care and Unit of Clinical Pharmacology and Geriatric Pharmacotherapy, School of Pharmacy, University of Eastern Finland, kuopio Campus, Yliopistonranta 1, P.O. Box 1627, FI-70211 Kuopio, Finland
  2. 2Institute of Public Health and Clinical Nutrition/Department of Geriatrics, University of Eastern Finland, Finland
  1. Correspondence to Sirpa Hartikainen
    Professor in Geriatric Pharmacotherapy, Kuopio Research Centre of Geriatric Care and Unit of Clinical Pharmacology and Geriatric Pharmacotherapy, School of Pharmacy, University of Eastern Finland, Kuopio Campus, Yliopistonranta 1, P.O. Box 1627, FI-70211 Kuopio, Finland; sirpa.hartikainen{at}

Statistics from

Commentary on:

Falls are a common geriatric syndrome. They are highly prevalent in older people, their impact on quality of life and disability is substantial, and multiple factors contribute to the risk.1 Age-related changes in pharmacokinetics and pharmacodynamics and declining health make older people particularly vulnerable to adverse drug events, including falls. Risk factors are of most importance in clinical practice if they are easily identifiable, modifiable and, at best, avoidable. Medication use often meets these criteria.

Two credible meta-analyses on drugs and falls were published by Leipzig and colleagues in 1999.2 3 Woolcott and colleagues recently provided a quantitative update to these meta-analyses. They demonstrated a significant association between the use of sedatives/hypnotics, antidepressants and benzodiazepines and falls in older people. Their meta-analysis was based on data from 22 original studies published between 1996 and 2007 and included a total of 79 081 participants aged 60+ years. None of the studies used a randomised controlled design. The meta-analysis included 5 case–control, 10 cohort and 7 cross-sectional studies, from 53 studies that were initially considered and assessed for inclusion. A clear description of inclusion and exclusion criteria is crucial to allow readers to interpret the results of any meta-analysis. Unfortunately, Woolcott and colleagues did not describe in detail their reasons for excluding 31 studies here. A study-by-study description of the reasons for exclusion would have reassured readers that all relevant studies were included.

The studies included in the meta-analysis were reviewed systematically with regard to definition of outcome (falls) and control of confounding factors, whereas the target medications were described only at the level of broad medication classes. The results were presented by nine medication classes (antihypertensives, diuretics, β-blockers, sedatives/hypnotics, neuroleptics/antipsychotics, antidepressants, benzodiazepines, narcotics and non-steroidal anti-inflammatory drugs). A description of the drugs or drug groups included in these classes would have been a welcome adjunct.

The use of Bayesian methodology will become more common in meta-analyses. Woolcott and colleagues described their use of this methodology as a key strength of their meta-analysis. One feature of the Bayesian approach is that it allows prior information to be integrated with new data. This allowed Woolcott and colleagues to build on evidence derived from the meta-analyses conducted by Leipzig and colleagues.2 3 However, the use of prior evidence can pose some problems. Although the same medication classes were used in the present and previous meta-analyses, the drugs within these classes have changed over time. For example, barbiturates, chloral hydrate and hydroxyzine are no longer used as sedatives; newer atypical antipsychotics have increasingly replaced phenothiazines and buturophenones propranolol is not among the most commonly used β-blockers; and tricyclic antidepressants have largely been replaced by selective serotonin reuptake inhibitors and other antidepressants. For this reason the Bayesian approach might have had a confounding effect on the results. On the other hand, one strength of this study was that the pooled risks were also calculated using frequentist methods. This facilitated risk comparisons between the ‘older’ and ‘newer’ medication classes.

Evidence from systematic reviews and meta-analyses should be used to guide clinical practice. This can present a challenge to researchers: their study design and interpretation of results should be adjusted to meet the needs of clinicians. Reporting results for antihypertensives, for example, does not meet these demands: the class is wide and unspecified, and grouping may dissipate potential risk differences among specific drug groups. Treating patients is about weighing risks and benefits, not just avoiding risks. Clinicians and patients need to know the risk profiles and appropriate dosages of specific preparations in order to select safe and effective treatments. Unfortunately, falls are not yet routinely considered as adverse drug events in clinical trials of drugs intended for older people. The current evidence on medication use and risk of falls is mainly based on observational studies, and many of these studies have methodological deficiencies.4 Thus, we need more high-quality studies concerning specific drugs and the risk of falls in older people. Systematic reviews and meta-analyses of these studies will help clinicians and patients select the most appropriate drug when pharmacotherapy is required.


View Abstract


  • Competing interests None.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.