Jonathan Brett

Objectives: To evaluate trends in pregabalin use and overdose in Australia and examine temporal relationships with media coverage (December 2018), prescribing and dispensing restrictions (April 2020), and boxed warnings (February 2021). Design: Retrospective observational study of dispensing and overdose data, with changepoint, interrupted time series, and latent class analysis. Setting and participants: Pharmaceutical Benefits Scheme data (PBS; 10% random sample of people with drug subsidies) and overdose calls (self-harm, recreational use, and misuse) reported to Australia’s largest Poisons Centre (50% national sample), 2016-24. Main outcome measures: Population adjusted rates of dispensing and overdose, timing of breakpoints and rate ratios, proportion of people who use pregabalin at high risk of misuse. Results: 227 864 people were dispensed 3 441 851 pregabalin prescriptions in the PBS 10% sample and 4871 pregabalin overdose calls were made to the Poisons Centre. Pregabalin use peaked in 2018 before plateauing and declining. Three breakpoints were found in dispensing data, with major decreases in level and slope: December 2018, May 2020, December 2021. The combined rate ratio for dispensing was 0.748 (95% confidence interval 0.700 to 0.800). Compared with the projected historical trend in the absence of interventions (counterfactual), approximately 118 390 (93 960 to 140 940) fewer pregabalin prescriptions were recorded nationally in December 2021, and an ongoing reduction of 0.7% (0.5% to 0.9%) was found each month. All breakpoints were associated with declines in overdose rates, with a combined rate ratio of 0.532 (95% CI 0.406 to 0.695). This reduction corresponded to approximately 41.3 (26.9 to 52.4) fewer pregabalin overdose calls in December 2021 and an ongoing reduction of 1.6% (1.0% to 2.3%) each month. Latent class analysis of 57 744 people dispensed pregabalin in the 12 months after the final breakpoint revealed four groups. The high risk group (7790, 13.5%) were more likely male, younger, dispensed more pregabalin, with shorter refill interval and high rates of opioids and benzodiazepine co-use (5120 (65.7%) and 2455 (31.5%), respectively). Conclusions: Pregabalin use and overdose in Australia have declined in recent years. Major decreases in trend corresponded with media coverage (December 2018) and prescribing and dispensing restrictions (May 2020). The third breakpoint may be a delayed impact of boxed warnings (implemented February 2021). These findings suggest that regulatory and risk communication measures may contribute to reducing pregabalin related harm.

Summary: Background: We used sequential target trial emulation (TTE) to examine associations between opioid analgesic initiation and adverse maternal and neonatal outcomes while accounting for gestational timing, time at risk, and potential confounding. Methods: Using the Early Life Course platform, comprising linked population-based data for births in New South Wales, Australia (2013–2019), we emulated a new trial at each gestational week, comparing women who initiated a prescribed opioid analgesic with eligible non-initiators. Baseline covariates were balanced using inverse probability weighting. Weighted logistic regression estimated odds ratios (ORs) for neonatal outcomes, postpartum haemorrhage, and severe maternal morbidity. Pooled logistic regression estimated hazard ratios (HRs) for other maternal outcomes, combined across gestational weeks using fixed-effects meta-analysis. Sensitivity analyses assessed unmeasured confounding, exposure misclassification, and potential reverse causation. Findings: Among 511,357 pregnancies, over 80% of opioid initiators received only one dispensing. Opioid initiation was not associated with increased risks of placental abruption (HR 0.94, 95% CI 0.75–1.16), severe maternal morbidity (OR 1.05, 0.97–1.11), postpartum haemorrhage (OR 0.98, 0.95–1.02), low Apgar score (OR 0.99, 0.87–1.08), small for gestational age (OR 0.90, 0.85–0.95), or neonatal abstinence syndrome (OR 0.98, 0.72–1.30). Risks were elevated for PPROM (HR 1.10, 0.98–1.23), preterm birth (HR 1.26, 1.19–1.33), severe neonatal morbidity (OR 1.08, 1.02–1.17), and perinatal death (OR 1.29, 1.10–1.52). Perinatal death estimates attenuated in analyses addressing potential reverse causation. Interpretation: Using a TTE approach that accounted for gestational timing and time at risk, we found no appreciable increase in most maternal or neonatal outcomes following predominantly short-term opioid initiation. Elevated risks of preterm birth, perinatal death, and third-trimester PPROM may reflect causal effects or confounding by pregnancy-related indications. Studies using time-appropriate designs and detailed information on prescribing indications and their timing are needed to clarify these associations. Research in Context: Evidence before this study: We searched the electronic databases MEDLINE (PubMed) for published articles, without language restrictions, from 1 January 2000 until 9 September 2026. The search strategy included terms for exposure (e.g. analgesic opioids), population (e.g. pregnancy and neonatal) and maternal and neonatal outcomes (e.g. preterm birth). Our search identified 10 observational studies which examined maternal and neonatal outcomes following opioid analgesic exposure during pregnancy. In these studies, prenatal opioid exposure above certain levels was consistently associated with neonatal abstinence syndrome (NAS). Beyond NAS, findings were inconsistent: some studies reported increased risks of peripartum haemorrhage, stillbirth, small for gestational age, low Apgar scores, preterm birth, and placental abruption, whereas others found no increased risk. It remains unclear whether observed associations are causal or reflect residual confounding. Methodological limitations may also have obscured associations, including immortal time bias when treatment initiation and follow-up were not aligned, and analyses that did not adequately account for gestational timing and therefore may have missed periods of greater susceptibility. As opioid analgesics are among the few options for managing pain during pregnancy, high-quality evidence on their safety is essential for informed clinical decision-making. Added value of this study: This study applied a sequential target trial emulation framework to estimate the effects of opioid analgesic initiation during pregnancy across a broad range of maternal and neonatal outcomes, aligning eligibility, treatment initiation, and follow-up while accounting for gestational timing. Opioid initiation was not associated with appreciably increased risks of most outcomes, including placental abruption, postpartum haemorrhage, severe maternal morbidity, and low Apgar scores. A small increase was observed for severe neonatal morbidity. Relative risks increased across trimesters for preterm birth, while PPROM and perinatal death were most elevated following third-trimester initiation. These findings may reflect opioid effects, although evolving pregnancy complications that both prompt opioid treatment and independently increase outcome risk remain a plausible explanation. NAS risk was not increased overall, although estimates were higher for oxycodone and higher oral morphine equivalent (OME) exposure. Implications of all available evidence: Taken together with broader evidence showing no substantial increase in the risk of major congenital malformations or adverse neurodevelopmental outcomes, these findings are generally reassuring for short-term opioid analgesic use in pregnancy, particularly codeine. The observed signals for preterm birth, PPROM and perinatal death, should not, in isolation, deter clinically indicated use, but support existing guidance to use opioids at the lowest effective dose for the shortest necessary duration and highlight the need for further studies that can distinguish direct opioid effects from shared clinical pathways.

Background: Alcohol, drugs and self-harm are important public health issues that often coincide with socioeconomic inequalities. While national policy frameworks highlight these as important issues, data describing their prevalence in unscheduled care are lacking. Aim: To conduct the first national-level description of the number, demographic characteristics and routes of care involving alcohol, drugs or self-harm emergencies across all unscheduled care services in Scotland. Methods: A retrospective cross-sectional study using linked data from all unscheduled care services in Scotland (NHS 24 telephone triage service; primary care out-of-hours; ambulance; emergency department (ED); acute admission and mental health admission). All continuous care episodes (defined as pathways) were analysed and within these, indicators for involvement of alcohol, drugs and/or self-harm were identified from service level ‘flags’ and clinical codes. Pathways from 1 April 2022 to 31 March 2023 were included. Descriptive analyses of characteristics, exploration of the top 30 most common pathways and top-down pathway-attributable-cost analyses were undertaken. Results: Of 2 721 120 pathways, 2.7% involved alcohol, drugs and/or self-harm—indicating considerable overlaps and complexity of care episodes. In 1.9% of all pathways only one of these indicators was identified. For the top 30 pathways (reflecting 79.0% of pathways involving only one indicator), the most common patient journey was ambulance to ED to acute admission, with an overall directly attributable cost to the National Health Service of approximately £216.9 million. Conclusion: Using current structured data, 1.9% of unscheduled care pathways in Scotland involve alcohol, drugs or self-harm and 2.7% involve one or several of these indicators. Ambulance services and EDs feature in most of these care pathways and are potential settings for intervention. Future work should explore these pathways in greater depth, refine case identification at service level and develop integrated care models within the ED to better serve this patient population and address health inequalities.

Objective: To measure the accuracy of clinical coding, clinician review, and a locally deployed large language model (LLM) in identifying alcohol, drug, and self-harm involvement in emergency department (ED) attendances, and quantify prevalence. Design: Two-phase diagnostic accuracy study. In a validation week, the identification strategies were assessed against a conflict-adjudicated reference standard (n=2,256); the LLM was then applied to n=105,096 annual attendances at the same site. Setting: UK Type 1 Emergency Department treating patients ≥16yrs. Main outcome measures: Prevalence quantification compared with the reference standard; sensitivity, specificity, and balanced accuracy of each strategy; monthly identification rates and adjusted annual prevalence. Results: The reference standard identified 12.1% of attendances as involving alcohol, drugs, or self-harm (coding 6.0%; clinician 10.0%, LLM 15.6%). LLM balanced accuracy matched or outperformed clinician review in all three domains (alcohol 0.942 v 0.930, p=0.635; drug 0.959 v 0.791, p<0.001; self-harm 0.982 v 0.908, p=0.004). Coding recorded 1.07 domains per identified patient against 1.32 in the reference standard. Adjusted annual prevalence corresponded to 12,890 domain involvements per year not identifiable in coded data. Subdomain classification found at least 81.6% of self-harm attendances required medical assessment for injury or overdose before psychiatric review. Conclusions: Clinical coding identified fewer than half of presentations involving alcohol, drugs, and self-harm and rarely captured co-occurring domains; under-recording was present across a full year. A locally deployed LLM generated more complete structured data from existing clinical text within NHS infrastructure, at a scale which is not feasible for manual review. What is already known on this topic: Coded emergency department data underpin national surveillance, commissioning, research, and service planning, but the size and direction of their error for specific conditions have not been measured against a validated reference standard in an unselected population. What this study adds: In a UK emergency department, clinical coding identified 6.0% of attendances as involving alcohol, drugs, or self-harm against an adjudicated reference-standard prevalence of 12.1%. In blinded head-to-head comparison, a locally deployed, open-source large language model achieved similar or higher balanced accuracy than clinician review. Applied to 105,096 attendances over 12 months, it identified 14.3% of attendances against 4.1% by coding, in every month of the year. How this study might affect research, practice, or policy: Prevalence estimates and service planning based on coded emergency department data are likely to understate both the burden and the co-occurrence of these conditions. The majority (81.6%) of classified self-harm presentations required medical assessment for injury or overdose; only 18.4% did not, a figure directly relevant to mental health emergency-care streaming. Locally deployed open-source language models offer a scalable, governance-compatible method for identifying these presentations from existing free text.

Background: Clinical trials comparing sustained-release (SR) tapentadol with controlled-release (CR) oxycodone have suggested that tapentadol (SR) may be associated with fewer nervous system side effects. However, it remains important to evaluate the relative safety of these agents in real-world clinical settings. Given the potential severity and economic burden of falls, together with their established association with opioid use, we compared the short-term risk of falls following initiation of tapentadol (SR) versus oxycodone (CR) in routine clinical practice. Methods and findings: We conducted an active comparator, new user retrospective cohort study using linked routinely collected health data on residents of New South Wales, Australia (2014–2020). We included people aged ≥18 years initiating publicly subsidised tapentadol (SR) or oxycodone (CR). Our outcome was a composite measure of fall-related emergency department presentations, hospitalisations or deaths. We used propensity score matching to adjust for baseline confounding and approximated relative risks (RR) of falls at 7, 14, and 28 days after initiation using conditional logistic regression models. We calculated absolute risk differences and estimated the number of people that would need to be treated with oxycodone (CR) versus tapentadol (SR) for one additional fall to occur (NNTH). We conducted subgroup analyses restricted to people aged ≥65 and ≥80 years and by recent opioid exposure (within 90 days prior to initiation). We identified 103,924 tapentadol (SR) and 419,732 oxycodone (CR) initiators; after matching each cohort comprised 103,758 initiators. Most people (74%) were aged between 45 and 84 years, and slightly more than half were female. Within 28 days of initiation, 652 (0.6%) oxycodone (CR) initiators and 457 (0.4%) tapentadol initiators experienced a fall. Across all time points, tapentadol (SR) initiation was associated with a lower risk of falls compared with oxycodone (CR) (7 days: RR 0.57 [95% CI: 0.48, 0.69]; 14 days: RR 0.62 [95% CI: 0.54, 0.71]; 28 days: RR 0.70 [95% CI: 0.62, 0.79]). Absolute differences were small at all time points (approximately 1–2 fewer falls per 1,000 patients treated), corresponding to NNTH for one additional fall ranging from 739 to 529. These patterns persisted regardless of recent opioid exposure. Relative risks were similar in the older age groups while absolute differences were slightly larger (≥65 years: 2–3 fewer falls/1,000; ≥80 years: 4–8 fewer falls/1,000). The greatest absolute differences were among opioid-naïve people aged ≥80 years (6–10 fewer falls/1,000, corresponding to NNTH ranging from 173 to 97). Fall risks were attenuated among people aged ≥80 years with recent opioid exposure. The main limitations of this study were that we did not have data on immediate-release tapentadol, or on falls that did not result in emergency department presentations or hospitalisations but may have otherwise impacted independence and mobility. Conclusions: Tapentadol (SR) was associated with a lower risk of falls than oxycodone (CR) up to four weeks after initiation, although absolute differences were small. The reduction in risk may be an important consideration in older patients where the consequences of falls are most severe.

The Urgent and Emergency Care system generates a wealth of clinical information, but our ability to harness this for public health planning and to address health inequalities is constrained by systemic data quality issues. Modern natural language processing (NLP), driven by the context-aware capabilities of transformer-based architectures and large language models, offers a transformative opportunity to bridge this gap. By training machines to interpret and structure context-rich clinical notes at scale, we can translate complex patient stories into data ready for research and systems intelligence that reflects the realities of real-world care. This technology offers a potential route to addressing health inequities in vulnerable populations, such as those presenting with crises related to mental ill-health, alcohol and drug use. Current reliance on structured but oversimplistic data often fails to capture the complex intersectionalities of clinical and social contexts. This is due to factors like diagnostic overshadowing and unrecorded multimorbidity, leaving these patients statistically obscured within routine datasets, which fail to accurately represent volume or complexity. This data invisibility perpetuates a cycle of inaccurate disease burden estimates, under-resourced services and flawed policy. By unlocking the detailed narrative data within unstructured notes, NLP could allow us to identify the acute social stressors and psychiatric contexts that are currently invisible, making these inequities visible and actionable.

Introduction: The objective of this study was to examine trends in opioid agonist treatment (OAT) during pregnancy for opioid dependence, by OAT type (methadone, buprenorphine) and during‐pregnancy OAT type change. Methods: Data linkage study. The cohort included pregnancies resulting in childbirth among people who received OAT within 4 years prior to birth, New South Wales, 1 January 2005–31 December 2021. We calculated the prevalence of OAT during pregnancy by OAT type and year of birth; prevalence rate ratios (PRR) and proportions experiencing OAT type change during pregnancy. Results: The cohort included 5212 childbirths among 3303 people. OAT was received in 4102 (78.7%) pregnancies. The annual prevalence of OAT during pregnancy was 77.2% in 2005, 86.2% in 2012 and 63.2% in 2021 (PRR 0.99; 95% confidence interval [CI] 0.99–1.00). Between 2005 and 2021, methadone use during pregnancy decreased from 70.9% to 37.4% (PRR 0.97; 95% CI 0.96–0.98) while buprenorphine use increased from 12.3% to 29.3% (PRR 1.07; 95% CI 1.06–1.08). Overall, changes in OAT type occurred in 226 pregnancies (5.5%). Among 971 pregnancies initially on buprenorphine, methadone was later used among 193, with this proportion decreasing over time. Of the 3131 pregnancies initially on methadone, buprenorphine was later used among only 33. Discussion and Conclusions: Three in four people in the cohort received OAT during pregnancy, with no statistically significant change in annual prevalence over the 2005–2021 period. Given prior evidence indicating higher rates of treatment discontinuation with buprenorphine versus methadone, the increased buprenorphine use during pregnancy highlights the need for targeted strategies to support treatment retention.

Background and aims: To measure the proportion of women in New South Wales, Australia, retained in opioid agonist treatment (OAT) for opioid dependence (OD) during pregnancy and examine how this varies according to treatment and maternal characteristics. Design: Retrospective cohort study using linked population‐based data, including OAT records, perinatal, hospital admissions, mental health outpatient services and criminal justice data. Setting: New South Wales, Australia, January 2004–December 2021. Participants: Pregnancies resulting in childbirth among women receiving OAT during pregnancy. The cohort included 4472 pregnancies among 2821 women receiving OAT during pregnancy. Measurements: We defined retention as continuous receipt of OAT from the date of OAT initiation during pregnancy or date of conception (whichever came last) until childbirth. We calculated the proportion of women retained in treatment and 95% confidence intervals (CI) overall, by timing of initiation (pre‐conception, first, second and third trimester) and medication type (methadone, buprenorphine) at initiation. We used logistic regression to assess retention variation according to maternal socio‐demographic and clinical factors, including morbidities commonly co‐occurring with OD, stratified by timing of initiation. Data on dosing, other substance use and psychopharmacological medications were unavailable. Findings: OAT was initiated pre‐conception in 74.8% (3346) of pregnancies, during the first trimester in 11.1% ( n = 497), second trimester in 8.8% ( n = 394) and third trimester in 5.3% ( n = 237). Overall, women were retained in OAT for 84.3% (3771) of all pregnancies. Retention was 87.4% (95% CI = 86.3–88.6) with pre‐conception initiation; 65.8% (95% CI = 61.6–70.0) in the first trimester; 80.5% (95% CI = 76.5–84.4) second trimester; and 85.2% (95% CI = 80.8–89.7) third trimester. Retention was 72.4% (95% CI = 69.5–75.3) with buprenorphine and 87.7% (95% CI = 86.6–88.9) with methadone. Pregnancies delivered during 2019–2021 were less likely to be retained in treatment compared with those delivered during 2004–2006, regardless of the timing of initiation [pre‐conception initiation odds ratio (OR) = 0.49, 95% CI = 0.30–0.79; first trimester initiation OR = 0.29, 95% CI = 0.08–1.07; second/third trimester initiation OR = 0.29, 95% CI = 0.10–0.91). Among women who initiated OAT pre‐conception, retention was lower among those whose first antenatal visit occurred after 20 gestational weeks (OR = 0.68, 95% CI = 0.53–0.86), those in their first (OR = 0.72, 95% CI = 0.54–0.97) or second pregnancy (OR = 0.73, 95% CI = 0.56–0.96) and those who initiated on buprenorphine (OR = 0.31, 95% CI = 0.24–0.40) or in a custodial setting (OR = 0.66, 95% CI = 0.44–1.01). Conclusions: In New South Wales, Australia, from 2004 to 2021, over 84% of women receiving opioid agonist treatment during pregnancy were retained in treatment until childbirth; however, lower retention in later study years and among women initiating buprenorphine, coupled with recent guidelines recommending buprenorphine as first‐line therapy during pregnancy, highlights the need for ongoing monitoring and targeted support to improve retention among women who are at higher risk of discontinuation.

Objective Given the potential severity and economic burden of falls, we compared the short-term risk of falls following initiation of sustained-release (SR) tapentadol versus controlled-release (CR) oxycodone in real-world clinical practice. Design Active comparator, new user retrospective cohort study using routinely collected health data from 2014 to 2020. Setting New South Wales, Australia. Participants People aged ≥18 years initiating publicly subsidised tapentadol (SR) or oxycodone (CR). Interventions Initiation of tapentadol (SR) or oxycodone (CR). Main outcome measures Composite measure of fall-related emergency department presentations, hospitalisations or deaths. We used propensity score matching to adjust for baseline confounding and approximated relative risks (RR) of falls at 7, 14, and 28 days after initiation using conditional logistic regression models. We calculated absolute risk differences and estimated the number of people that would need to be treated with oxycodone (CR) vs tapentadol (SR) for one additional fall to occur (NNTH). We conducted subgroup analyses restricted to people aged ≥65 and ≥80 years and by recent opioid exposure (within 90 days prior to initiation). Results We identified 103,924 tapentadol (SR) and 419,732 oxycodone (CR) initiators; after matching each cohort comprised 103,758 initiators. Most people (74%) were aged between 45-84 years, and slightly more than half were female. Within 28 days of initiation, 652 (0.6%) oxycodone (CR) initiators and 457 (0.4%) tapentadol initiators suffered a fall. Across all time points, tapentadol (SR) initiation was associated with a lower risk of falls compared with oxycodone (CR) (7 days: RR 0.57 [95% CI 0.48 to 0.69]; 14 days: 0.62 [0.54 to 0.71]; 28 days: 0.70 [0.62 to 0.79]). Absolute differences were small at all time points (approximately 1-2 fewer falls per 1,000 patients treated), corresponding to NNTH for one additional fall ranging from 739 to 529. These patterns persisted regardless of recent opioid exposure. Relative risks were similar in the older age groups while absolute differences were slightly larger (≥65 years: 2-3 fewer falls/1,000; ≥80 years: 4-8 fewer falls/1,000). The greatest absolute differences were among opioid-naïve people aged ≥80 years (6-10 fewer falls/1,000, corresponding to NNTH ranging from 173 to 97). Fall risks were attenuated among people aged ≥80 years with recent opioid exposure. Conclusions Tapentadol (SR) was associated with a lower risk of falls than oxycodone (CR) up to four weeks after initiation, although absolute differences were small. The reduction in risk may be an important consideration in older patients where the consequences of falls are most severe.

Background and aims: There is an urgent unmet need for novel treatments for methamphetamine (MA) use disorder. We explored the qualitative experiences of people participating in a study of psilocybin‐assisted psychotherapy (PAT) to treat MA use disorder. Design and setting: Qualitative study of participants enrolled in a single arm, open‐label pilot study of PAT for MA use disorder delivered in an outpatient stimulant treatment program setting in Sydney, Australia. Participants: Twelve participants were interviewed before starting PAT and then again one month following PAT. Measurements: Pre‐PAT interviews explored participants' experiences of MA use and expectations of receiving PAT. Post‐PAT interviews explored participants' experiences of PAT, with a focus on phenomena related to the acute subjective effects of psilocybin, the perceived effects of PAT on MA use, self‐perception, beliefs, values, behaviours, interpersonal relationships and spirituality, and acceptability of the intervention. Interviews were audio recorded, transcribed verbatim and analysed using an inductive qualitative approach. Findings: While participants generally hoped to have positive outcomes from study participation, their expectations were generally tempered and realistic. Their trial experiences of PAT were often characterised by new understandings of themselves, their narrative histories and interpersonal relationships, all of which were frequently prompted by leaning into vividly presented challenging experiences within the psychedelic experience. This volitional attitude of ‘leaning into the obstacle’ emerged as a key theme, meriting exploration for its potential to expose the subjective dimension of the psychedelic mechanism of effect. Resolution of this obstacle was associated with a reduction in the salience of methamphetamine. Therapeutic alliance was seen as critical to positive outcomes and was achieved through high levels of concentrated therapeutic attention and intersubjective intimacy between participant and therapist. Conclusions: Interviewed participants in a study of psilocybin‐assisted psychotherapy (PAT) for methamphetamine use disorder perceived PAT as an acceptable intervention. Transformation in understandings of self and interpersonal relationships and subsequent reduced salience of methamphetamine use often occurred through confronting psychic obstacles in the context of high levels of therapeutic support from study therapists.