Background. Previous research with samples recruited in primary care and drug treatment has shown that opioid agonist therapy (OAT) reduces injection drug use frequency. We sought to model transitions between states of drug injecting and non-injecting over time and the impact of OAT on these transitions in a community-based cohort of people who inject drugs.

Methods. Data come from SuperMIX, an ongoing prospective cohort study involving annual interviews with people who inject drugs in Melbourne. Directed Acyclic Graphs established a data-generating process for a causal effect between OAT engagement and cessation of any injecting drug use between annual interviews. Using random-effects multi-state modelling, we modelled transitions between states of injecting and non-injecting to estimate the effects of OAT engagement on these transitions.

Results. From 1496 participants with 4875 person-period follow-up (FU) observations, we observed 281 injecting cessation and 139 injecting resumption events. Preliminary analyses provide evidence of a time-dependent positive association between any OAT engagement and injecting cessation (adjusted Hazard Ratio [aHR]=2.29, 95% confidence interval [95%CI] = 1.08, 4.84 at 7-years follow up). Similarly, sustained engagement in OAT increased likelihood of cessation (aHR=1.12, 95%CI=0.99, 1.26). Any OAT was found to increase injecting cessation among those who identified as Aboriginal or Torres Strait Islander (aHR=5.99, 95%CI=1.07, 33.6). There was a small to moderate correlation (r=.28) between each of the state-specific random effects indicating participants were most likely to belong to one of two groups: frequent-movers (high probability for both states – short periods of cessation and resumption) or non-movers (low probability for both states – longer periods of cessation and resumption).

Conclusion. We found some evidence of a time-dependent causal relationship between OAT engagement and cessation of injecting drug use which, importantly, reduces the likelihood of drug-related harm in a community-based cohort of people who inject drugs.

Background: Unmeasured confounding is major concern in pharmacoepidemiological research. Instrumental variable (IV) analyses offer a potential solution. We sought to map the contemporary IV application in pharmacoepidemiological research. Our objective was to describe: 1) the frequency and characteristics of studies in prescription medicine research that employ IV analyses, 2) the implementation of IV analyses, and 3) whether IV validity was demonstrated.

Methods: We searched MEDLINE, Embase, PsycINFO, CINAHL Complete, and National Bureau of Economic Research databases, for studies that used an IV to assess the safety and/or effectiveness of medicines (including vaccines) published between January 1, 2000 and August 1, 2022. We included studies where the medicine was either the intervention or control; or where the same medicine was being evaluated at different doses, routes, or formulations. We extracted the target IV, its operationalisation, its role (primary/sensitivity analysis), and any evidence for the relevance, exclusion, and exchangeability assumptions. The study protocol was registered with Open Science Framework (DOI: 10.17605/OSF.IO/V96MG).

Results: Of the 4,372 studies screened, 202 met the eligibility criteria, providing information on 253 IVs. Almost half (90/202) of all IV studies were published in the last five years. The US and Japan accounted for 74% (149/202) of all studies. IVs increasingly serve as sensitivity analyses to validate the robustness of the main analyses. The most common IV was facility preference (100/252), followed by physician preference (50/252), calendar time (26/252) and regional variation (24/252). Only 31.7% (64/202) studies explicitly reported on all three IV assumptions.

Conclusions: IV analysis is being increasingly used in PE research, and as a sensitivity analysis with preference-based IVs being most commonly used. Infrequent reporting of tested assumptions is potentially cause for concern and researchers should improve reporting to ensure robust interpretation of findings.

Background: The objective of this study was to assess whether prenatal exposure to attention-deficit/hyperactivity disorder (ADHD) medication increases the risk of neurodevelopmental disorders in children.
Method: Our population-based cohort study used nationwide register data on liveborn children from singleton pregnancies (2005-2020) from Norway (NO) and Sweden (SE). We compared use of ADHD medications during pregnancy, stratified by period of exposure and individual medication. We used propensity score-weighted Cox proportional hazard regression to estimate adjusted hazard ratios (aHRs) and 95% confidence intervals (CIs) for risk of neurodevelopmental disorders (ICD10 codes: F70-F73, F80-F89, F90-F98).
Result: Among 2,384,704 children (median follow-up time: 6 (SE) & 8 (NO) years), 5610 (0.2%) were exposed to ADHD medication prenatally. We observed elevated risk estimates for any neurodevelopmental disorder when considering; all ADHD medications (aHR: 1.19, 95% CI 1.06-1.34) & methylphenidate exposure (aHR: 1.16, 95% CI 1.02-1.32). However, results were not consistent across secondary analyses, we observed no effect in crude or adjusted risk estimates when restricting to pregnant women with ADHD: aHR of 1.03 (0.92-1.16), and when comparing use with those who discontinued treatment prior to pregnancy: aHR of 0.99 (0.87-1.13) and in a sibling analysis aHR of 1.00 (0.82-1.23). Results were consistent across individual neurodevelopmental disorders, with the exception of tic disorders which showed elevated risk estimates across several analyses.
Conclusion: The findings of this multinational cohort study, align with previous research, and suggest there is little to no increased risk of child neurodevelopmental disorders after prenatal exposure to ADHD medication. Elevated effect estimates observed in the overall population were likely due to unmeasured confounding, despite access to detailed health information. Our findings strengthen existing evidence and can assist pregnant people and clinicians in weighing the benefits and risks of medication use during pregnancy.

Background: In light of recent measures to curb opioid use and harms, dispensings of opioid analgesics for pain, subsidised under Australia’s Pharmaceutical Benefits Scheme (PBS), have been declining, but information on trends in non-PBS subsidised use (private market and public hospital) is not readily available. This study used two real world data (RWD) sources to describe 8-year population-level trends in Australia's opioid analgesic use.

Methods: Our descriptive study used two datasets covering 2015─2022: national IQVIA data on all (PBS/private) pharmaceutical sales to community pharmacies, hospitals and other settings, and PBS dispensing data for a 10% sample of Australian residents, extrapolated to estimate national PBS use. We measured total units of each opioid sold/dispensed, converted into oral morphine equivalent per 1000 population per day (OME/1000/day). We calculated non-PBS use by subtracting PBS OME dispensed from total OME sold. Hospital OME were calculated using IQVIA data on sales to hospitals.

Results: PBS dispensing data accounted for 86.2% of prescription opioid analgesic use in 2015, decreasing to 73.0% in 2022. Between 2015─2022 total opioid use decreased by 21.2%, from 1231.4 to 970.6 OME/1000/day. There were contrasting trends across PBS versus non-PBS use: between 2015─2022, PBS use decreased by 33.3%, or -353.4 OME/1000/day, from 1061.7 to 708.4 OME/1000/day, whereas non-PBS use increased by 55.4%, or +92.5 OME/1000/day, from 169.7 to 262.3 OME/1000/day. Opioid use in hospitals (public/private) remained stable, accounting for 8─10% of total use between 2015─2022.

Conclusion: This study provides real-world evidence (RWE) on opioid trends, in light of measures to reduce the use of opioids in Australia. Total opioid analgesic use declined between 2015─2022, but private dispensings, where the patient pays the full cost, increased substantially. Relying solely on PBS dispensing data for monitoring prescribed opioid trends may not be appropriate due to the increasingly significant under-ascertainment of total opioid analgesic use.

Background
Real-world evidence plays an integral role in understanding quality use of medicines in routine clinical care. SGLT2i are a pillar of therapy for people with type 2 diabetes, but recommendations to withhold temporarily during hospitalisation (due to rare risk of ketoacidosis) raise concern of inadvertent discontinuation following discharge. We used population-level data to examine if hospitalisation is a risk factor for SGLT2i discontinuation, and if this differed to other diabetes therapy (DPP-4i) not typically withheld in hospital.

Methods
We conducted a retrospective cohort study using linked population-level hospital, dispensing claims and mortality data for all adult residents of New South Wales. We identified two cohorts of adults aged 40+ years initiating either SGLT2i or DPP-4i between 2016 to 2020. We measured persistent use of therapy, based on contiguous periods of estimated medicine supply. We used Cox proportional hazards models to estimate adjusted hazard ratios (HR) for discontinuation, using time-dependent exposures of having a recent hospitalisation (during, and/or the 90 days following discharge) during follow-up.

Results
Of people initiating SGLT2i (n=106,098, median age 63, 61% male), 42.6% were hospitalised and 63.1% discontinued SGLT2i during follow-up. Discontinuation rates were higher in periods following a recent hospitalisation (16.7 per 10,000 person-years) compared to periods without a recent hospitalisation (10.1 per 10,000 person-years). This association remained in adjusted models (HR 1.63; 95% CI 1.59-1.66). Of people initiating DPP-4i (n=91,960, median age 66, 57% male), 47.6% were hospitalised, and 64.0% discontinued DPP-4i during follow-up. Discontinuation rates were also higher around periods of hospitalisation (HR 1.40, 95% CI 1.37-1.43).

Conclusion
We found hospitalisation was a risk factor for discontinuation of diabetes medicines. Our real-world evidence highlights a need for targeted communications to increase clinician and patient knowledge about the importance for continued SGLT2i use, as well as broader maintenance of diabetes therapy following discharge.

Background: The prevalence of multiple sclerosis (MS) has increased in recent decades. Using our medications estimation method, we aimed to calculate MS prevalence in Australia for 2024 as we have done previously in 2010, 2017, and 2021. Updated prevalence estimates will facilitate analyses of social and economic outcomes related to the disease, including a cost of illness study by our group. They will also support advocates of increased funding for MS research and interventions.

Methods: Following our validated and rapid medications methodology, Australia prescription data from the Pharmaceutical Benefits Schedule item reports were used to estimate the number of Australians with MS using MS disease-modifying therapies (DMTs). Subsequently, we extrapolated the number of persons with MS not using MS DMTs through the utilisation of DMT penetrance rates that were based on data from the Australian Multiple Sclerosis Longitudinal Study. Prevalence per 100,000 was calculated using the most recent Australian Bureau of Statistics population data.

Results: An estimated 37,756 people were living with MS in Australia in 2024, up 77% (16,442) since 2010. Over the same period, crude prevalence per 100,000 Australians also rose by 45.5% (or 43.5 persons) to 139.2/100,000. Tasmania retained the highest prevalence among the Australia states with an estimated age-standardised prevalence of 190.1/100,000 (95% CI: 188.5-191.8). Queensland had the lowest estimated age-standardised prevalence at 99.8/100,000 (95% CI: 98.6-100.9). This is demonstrative of a latitudinal gradient in MS prevalence. Of the 37,756 Australians living with MS in 2024, 62% were identified to be using MS DMTs. The most commonly used DMT was ocrelizumab, employed in 31% of treated cases.

Conclusions: Consistent with global trends, MS prevalence continues to escalate in Australia. This is attributable to both improved survival among people living with MS, better diagnosis, and an increasing rate of MS incidence.

Background
Pregnant women with opioid dependence are recommended to initiate and/or maintain opioid agonist treatment (OAT). In Australia, methadone has traditionally been the first-line treatment, while guidelines have endorsed buprenorphine for prenatal use since the mid-2010s. Monitoring OAT use during pregnancy helps identify treatment gaps and inform interventions. We assessed the annual prevalence of OAT use during pregnancy and OAT type switching among women with a history of opioid dependence treated with OAT
Methods
We linked OAT prescription authority data to records of pregnancies resulting in childbirth in New South Wales (2005-2021). Our cohort comprised women who used OAT in the four years prior to giving birth. We defined OAT use during pregnancy as prescription authority duration overlapping the gestation period. We calculated prevalence and annual change in OAT use during pregnancy (prevalence rate ratios, PRR), and proportions switching OAT types.
Results
Our cohort included 3303 women with 5212 pregnancies resulting in childbirth between 2005 and 2021. OAT was used in 4102 (78.7%) pregnancies. Annual prevalence of OAT did not change significantly (77.2% in 2005, 86.2% in 2012, 63.2% in 2021, PRR 0.99, 95% CI 0.99-1.00). Between 2005 and 2021, methadone use 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, switching OAT types occurred in 226 pregnancies (5.5%). Among 900 pregnancies initially on buprenorphine, 122 switched to methadone, proportionally decreasing over time. Among 3202 pregnancies initially on methadone, 104 switched to buprenorphine, with the proportion remaining consistent.
Conclusion
It is reassuring that three in four women in the cohort used OAT during pregnancy, with most maintaining their treatment choice. The increasing use of buprenorphine highlights the need for strategies to support opioid-dependent pregnant women in maintaining treatment given existing evidence of suboptimal retention for buprenorphine.

Introduction:
Evidence on opioid analgesic safety during pregnancy remains inconclusive. We employed a sequential target trial emulation (TTE) to generate real-world evidence on the risks of adverse maternal and neonatal outcomes following opioid analgesic exposure during pregnancy.
Methods:
We linked records of all pregnancies resulting in a birth in NSW (2013-2019) to prescription medicine dispensing, hospital admission, and mortality data. Sequential trials were created during each gestational week, assigning eligible women as either opioid initiators or non-initiators based on whether they were dispensed an opioid in that week. We employed inverse probability weighting to balance baseline covariates for each trial and incorporated treatment weights in logistic regression models to estimate odds ratio for neonatal outcomes, postpartum heamorrhage, and severe maternal morbidity, and pooled logisitic regression to estimate the hazard ratios for the remaining maternal outcomes. Bootstrapping was used to calculate 95% confidence intervals (CI).
Results:
The TTE yieled 18,360,776 trials (511,357 pregnancies; 398,944 women), consisting of 26,984 opioid initiators and 18,333,792 non-initiators. Comparing to non-initiators, the risk estimates following opioid initiation were 1.12 (95% CI 1.00–1.26) for preterm premature rupture of membranes , 1.27 (95% CI 1.20–1.35) for preterm birth,1.05 (95% CI 0.99–1.12) for severe maternal morbidity, 1.08 (95% CI 1.00–1.17) for severe neonatal morbidity, 1.28 (95% CI 1.04–1.59) for stillbirth, and 1.29 (95% CI 0.94–1.79) for neonatal death. No increase in risk was observed for placental abruption, postpartum hemorrhage, low Apgar score, small for gestational age or neonatal abstinence syndrome.
Conclusions:
Our findings suggest a slight increase in risk for certain maternal and neonatal outcomes. However, further research is needed to determine whether these risks are attributable to opioid initiation or are a result of reverse causation or residual confounding. Our TTE approach effectively mitigated common design-induced biases, particularly immortal time bias.