Deprescribing in Older People Approaching End of Life

A Randomized Controlled Trial Using STOPPFrail Criteria

Denis Curtin, MB; Emma Jennings, MB; Ruth Daunt, MB; Sara Curtin, MSc; Mary Randles, MB; Paul Gallagher, PhD; Denis O'Mahony, MD


J Am Geriatr Soc. 2020;68(4):762-769. 

In This Article


In this study of very frail older hospitalized patients with limited life expectancy, application of STOPPFrail criteria at a single time point resulted in a sustained and significant reduction in polypharmacy and average aggregate monthly medication costs at 3 months after randomization compared with usual pharmaceutical care. We found that almost 1 in 4 medications were discontinued in frail older people with polypharmacy using this method, resulting in a 28% average reduction in monthly medication costs. No significant differences were found between the intervention and control arms in terms of important health-related outcomes including unplanned hospital admissions, falls, fractures, QoL, and mortality, although it must be acknowledged that the trial was likely to have been underpowered to detect differences in these secondary outcomes.

Other structured deprescribing methods have recently been evaluated in very frail older people using a randomized controlled trial design, and they also reported statistically significant reductions in potentially inappropriate prescriptions. Potter et al[29] used an implicit (ie, judgment-based) algorithm that requires the user to answer a series of questions about each drug in the patient's regimen; Wouters et al[30] evaluated the Multidisciplinary Multistep Medication Review. Both methods are patient centered and comprehensive but limited by a requirement for resource-intensive processes that may hinder their integration into routine clinical practice. STOPPFrail overcomes these limitations by virtue of its conciseness and high interrater reliability between users of different disciplines and professional grades.[15,16]

The most common reason for deprescribing in this trial was when a drug had no clear valid clinical indication (STOPPFrail criterion A2). We contend that routinely clarifying whether a drug is linked to a verifiable diagnosis or an active or recurring symptom is fundamental to any formal medication review in older multimorbid patients exposed to polypharmacy, particularly frailer patients with very limited survival prospects. The remaining criteria in STOPPFrail are predominantly explicit and target specific drugs that, under the usual circumstances, may be clinically indicated but are likely to be associated with negligible benefits or net harm in the context of advanced irreversible frailty and limited life expectancy. During the conduct of the trial, it became clear that some of the explicit criteria in STOPPFrail lacked clinical relevance and were very seldom, if at all, applied (eg, systemic estrogens for menopausal symptoms, selective estrogen receptor modulators for osteoporosis). Likewise, it was evident that some medications, commonly prescribed in frail older people but lacking a firm evidence base (eg, vitamin D supplements), were absent from STOPPFrail. Future iterations of the criteria will aim to address these shortcomings.

Our study has several important limitations. First, we enrolled participants from just two acute hospitals in one city in Ireland which may limit the generalizability of our findings. STOPPFrail criteria were developed in the university affiliated with these hospitals, and this may have influenced the readiness of attending physicians to implement the deprescribing recommendations. Second, the attending physicians and participants were not blinded to intervention or control group allocation. Although this had the potential to introduce bias, we believed that, given the nature of the intervention, blinding would have been inappropriate. The research physician who identified deprescribing targets using STOPPFrail was also unblinded to the group allocation of participants, which, in theory, could have influenced the nature of the intervention. Furthermore, there were no quality control measures to assess the accuracy of the STOPPFrail-guided deprescribing recommendations. However, STOPPFrail criteria are predominantly explicit, and this, in effect, would be expected to limit variability in their deployment. Nonetheless, even though the trial was unblinded, the measured outcomes (apart from QoL assessments) were not subject to bias. Third, we were unable to determine the effect of the intervention on important patient-related outcomes including mortality due to the relatively small sample size. Fourth, because of resource restrictions, it was not possible to actively collect data on adverse drug withdrawal events or disease relapses due to the deprescribing intervention. Consequently, we may have missed these events if they were not reported to the study team. Finally, we did not use a cluster randomization design that would diminish the possibility of contamination bias. Physicians may have simultaneously had both intervention and control patients under their care during the trial and, through a "training effect," they may have applied STOPPFrail criteria during medication reviews of control patients. However, any possible contamination of this kind would increase the chance of actual effects of the intervention not being detected (ie, type II error). In spite of the possible presence of contamination, significantly different effects of the STOPPFrail intervention were still observed between the groups.

This study also has notable strengths. We included a representative sample of real-world highly frail, multimorbid older people, approximately 75% of whom had a known diagnosis of dementia. This patient group are commonly excluded from clinical trials despite being encountered with increasing frequency in clinical settings and having the highest levels of disease burden.[31] Deprescribing recommendations for this vulnerable population were implemented under medical supervision in the acute hospital and only after review and approval by participants' attending physicians. In addition, the deprescribing intervention in this trial was relatively straightforward and would be easy to replicate in other settings.

When very frail older people approach end of life, the prescription of multiple medications may be burdensome or even futile in their clinical management. This exploratory study provides evidence that STOPPFrail, an easily applied reliable and explicit deprescribing tool, substantially reduces polypharmacy and monthly medication costs in this patient cohort. Although there were no differences between groups for important clinical end points such as unscheduled hospital presentations, QoL, and mortality, the trial was very likely underpowered to detect significant changes in these outcomes. A larger scale multicenter trial with greater statistical power is required to reassure clinicians that STOPPFrail-guided deprescribing of long-term medications can be achieved in the frailest older people without compromising clinical outcomes.