Spotlight commentary: Addressing the persistent problem of benzodiazepine overuse, time for clinical, educational and AI‐driven reform

A recent article in the BJCP by McCarthy et al. once again underscored the fact that benzodiazepines (BZDs) are one of the most prescribed medications overall. This study analysed the prescription patterns of general practitioners in Ireland using The Drug Utilization 90% Index (DU90%), an index representing the number of medicines making up 90% of a doctor's prescribing, showed that almost 4% of all prescribed drugs are BZDs no matter the specific presciber-level characteristics.1 They found that a higher DU90% was positively asocciated with higher BZD presription rates.1 This results are in line with the notion that BZD prescribing has risen globally and consequently raised public health concerns worldwide (Figure 1). When looking at available data, in the United States, the share of adults prescribed BZDs grew from 4.1% (1996) to 5.6% (2013), while simultaneously overdose deaths involving BZDs increased from 0.58 to 3.07 per 100 000.2 The 2015–2016 National Survey on Drug Use and Health (NSDUH) found that 30.6 million adults (12.6%) used BZDs in the past year: 25.3 million (10.4%) used them as prescribed, while the use of BZDs in 5.3 million (2.2%) was classified as misuse.2 Furthermore, the prescribed use was highest in ages 50–64 (12.9%); misuse was highest in ages 18–25 (5.2%), often co-occurring with nonmedical opioid or stimulant use, while the most common way of misuse was obtaining BZDs without a prescription from friends or family.2 A survey conducted by Hockenhull et al. in England, and published in the BJCP, found that the past-12-month nonmedical use (NMU) of any BZD or Z-drug was 1.2% (~635 000 adults), of which 0.7% (~350 000) was for only BZDs and 0.5% (~250 000) was for only Z-drugs. England also shows large prescription volumes (≈1.4 million adults prescribed BZDs and ≈1.0 million prescribed Z-drugs in 2017–2018), raising diversion risk.3 NMU was higher in younger adults (≤34 years) and more common in women. The NMU was strongly associated with recent illicit drug use: stimulant users had the highest BZD NMU (cocaine 5.9%, amphetamine 5.6%, MDMA 5.2%). Again, a concerning trend was that BZDs were most commonly acquired via family or friends and then via dealers or online.3 Frequently prescribed to patients with preexisting conditions, BZDs place them at elevated risk for adverse drug events, including chronic obstructive pulmonary disease (COPD), substance use disorders and osteoporosis. A large-scale longitudinal cohort study involving 65912 patients in a primary care practice-based research network found that 15% of patients received at least one BZDs prescription within a 1-year period, with 5% of those receiving high doses.4 Compared to nonusers, BZDs recipients were significantly more likely to have comorbidities such as depression, substance or alcohol abuse, tobacco use, osteoporosis, COPD, sleep apnoea and asthma. For example, BZD use was associated with nearly triple the odds of depression and more than double the odds of substance abuse. High-dose recipients showed particularly pronounced associations, with substance abuse risk exceeding sevenfold risk compared to nonusers. In addition to these clinical risks, BZD users had greater healthcare utilization, with more outpatient visits, emergency department visits and hospitalizations.4 Moreover, the use of BZDs for sedation in intensive care units (ICUs) has been strongly associated with an increased risk of delirium compared to alternative sedatives such as propofol. A prospective cohort study conducted by van Gelder et al. in the BJCP, involving 950 ICU patients, assessed the incidence of delirium within 7 days following continuous deep sedation with either midazolam, propofol or a combination of both.5 Overall, 64% of patients developed delirium during this period. Notably, the incidence was highest among those who received midazolam alone, with 73% experiencing delirium. Patients who received both midazolam and propofol had a slightly lower but still elevated incidence (68%), while those sedated exclusively with propofol exhibited the lowest rate (54%). Statistical analysis further confirmed that midazolam sedation was independently associated with a significantly increased risk of delirium (adjusted cause-specific hazard ratio HR = 1.32), compared to propofol.5 The combination of propofol and midazolam was also associated with an elevated risk (HR = 1.29).5 Recent research by Zhou et al. in the BJCP indicates that chronic BZDs use disorder (BUD) is associated with notable structural brain changes and cognitive decline, particularly involving the hippocampus and the brain's waste clearance system, known as the glymphatic system.6 This imaging study has shown that individuals with BUD exhibit enlarged choroid plexus volumes and reduced volumes in several hippocampal subregions when compared to both healthy individuals and those with chronic insomnia. These changes were specific to BUD and not observed in insomnia alone, suggesting a direct effect of prolonged BZDs use.6 Furthermore, among those affected, increased choroid plexus size was linked to both hippocampal shrinkage and impaired glymphatic function. Cognitive assessments revealed that reductions in hippocampal volume were associated with poorer performance, indicating that structural brain alterations in BUD may contribute to cognitive impairment.6 Numerous studies have shown that drug interactions further magnify the dangers surrounding BZD use, most notably with opioids. In a cohort study of US veterans, Park et al. showed that, from 2004–2009, veterans who received opioid analgesics were prescribed BZDs in 27% of cases. Further analysis demonstrated that BZD use, whether current or past, was associated with a substantially increased risk of overdose.7 Sun et al., by analysing retrospective claims data from 2001 to 2013, found that the concurrent use of opioids and BZDs rose from 9% to 17% during the observed time period.8 Compared with opioid-only users, concurrent use was linked to greater odds of an emergency room visit or inpatient admission for opioid overdose.8 These findings are even more concerning if we take into account the emergence of designer or novel BZDs on the illicit market (e.g., etizolam and phenazepam). These compounds are increasingly available online or via diversion, may be more potent than standard prescription BZDs and have been implicated in a number of drug-related deaths (notably in Scotland). The unpredictable potency and composition of illicitly sourced BZDs amplify the risk of respiratory depression, overdose and death when these agents are used alone or combined with opioids or other depressants.3 Several studies in the BJCP demonstrated that interactions with CYP3A4 inhibitors can significantly intensify the effects of BZDs. For instance, Matsuki et al. showed that the CYP3A4 inhibitor Carotegrast methyl, an ulcerative colitis treatment, substantially increased exposure to midazolam, a commonly used BZD.9 In healthy volunteers, co-administration led to nearly double the peak concentration and more than triple the overall exposure to midazolam, indicating a marked interaction.9 Likewise, Knöchel et al. displayed that Atuliflapon, an asthma treatment, when co-administered with midazolam, increased the area under curve (AUC) by 56%.10 These findings demonstrate the potential for clinically significant drug–drug interactions, underscoring the need for dose adjustments or close monitoring when CYP3A4 inhibitors are used alongside BZDs. Despite these well-established risks, BZDs prescribing remains high and often inappropriate. Several recent articles published in the BJCP have directly examined the prescribing of BZDs and the effects of policy and regulatory interventions. Bennie et al. quantified polypharmacy among older Europeans across six countries, finding substantial between-country variation. In four countries, more than 50% patients had more than five prescriptions, while BZDs showed the highest variation among observed drugs, spanning from 2.7% in the UK to 34.9% in Spain, stressing the need for better polypharmacy management in Europe.11 Couret et al., using the FRA-DEM cohort, evaluated the impact of modifications to anti-dementia drug reimbursement in France and found that the policy change was followed by an increase in BZD prescribing, illustrating a critical and unintended consequence of reimbursement decisions for patient safety.12 Finally, Son and Je analysed South Korea's Narcotics Information Management System to evaluate whether the regulatory intervention reduced inappropriate prescriptions of BZDs. They found no significant improvement, which suggests that regulatory measures alone may be insufficient and that complementary strategies, for example, enhanced prescriber education and active monitoring, are needed.13 Recent studies by Bužančić et al. in the BJCP explored artificial intelligence (AI)-assisted BZDs deprescribing to underscore the potential for innovative solutions. In a real-world Croatian cohort of 154 BZD patients, healthcare providers and ChatGPT-4 applied pre-established deprescribing criteria, showing substantial, although not perfect, agreement (95% agreement rate).14 AI was able to identify a high proportion of patients eligible for deprescribing (89.6% vs. 96.1% by clinicians) and demonstrated agreement across specific criteria ranging from 74.7% to 91.3%, suggesting its potential to standardize assessments and flag opportunities that might otherwise be overlooked. However, reliance exclusively on AI recommendations remains problematic, as 22.1% of outputs were ambiguous, overly generic or inaccurate, posing risks for inappropriate decision-making.14 Thus, AI should be seen as a supportive tool rather than a replacement for clinical judgement. A recent scoping review by Okati et al., published in the BJCP, has shown that deprescribing apps with AI features are available on the market and that there is an evident need for a personalized approach, but underscores the need for further investigation to use them as an independent clinical tool.15 While BZDs remain a valuable therapeutic option, especially in acute states and as a short-term therapeutic option, their well-documented risks necessitate caution when prescribing. Greater adherence to evidence-based guidelines, improved clinician education and targeted policy interventions have become essential for safer use of BZDs. Furthermore, the use of novel AI algorithms can help address several pressing issues, including identifying at-risk patients and prescribers through predictive analytics, utilizing natural language processing to flag abnormal prescription patterns and detect misuse in real-time and employing automated auditing to detect over-prescribing. It is evident that this issue will require multidisciplinary cooperation, evidence-driven strategies to reduce overprescribing and innovations such as AI-assisted decision-making in order to be resolved effectively. Open access publishing facilitated by Sveuciliste u Zagrebu, as part of the Wiley - National and University Library in Zagreb Consortium Croatian Academic and Research Libraries Consortium agreement.