The impact on poisonings of up‐scheduling of modified release paracetamol to Schedule 3 (pharmacist only medicine)

n Australia, paracetamol is the agent most frequently implicated in drug overdoses, and their frequency is increasing, particularly in young people. 1 Paracetamol overdose causes significant morbidity despite treatment, and is the leading cause of acute liver failure in Western countries. 2 Modified release (MR) paracetamol overdose is associated with a higher rate of liver injury than immediate release paracetamol. 3 The sole therapeutic benefit of MR paracetamol is its more convenient dosage regimen (three rather than four times a day). 4 In response to the rising numbers of overdoses, the Therapeutic Goods Administration (TGA) up- scheduled MR paracetamol, from Schedule 2 to Schedule 3, in June 2020. 5 A pharmacist must be involved in sales of Schedule 3 medicines, which must be stored behind the pharmacy counter. We evaluated whether re- scheduling was associated with changes in the numbers of overdoses with MR paracetamol, immediate release paracetamol, and other over- the- counter analgesics reported to the New South Wales Poisons Information Centre (NSWPIC) to the end of August 2022. The study was approved by the Sydney Children’s Hospitals Network Human Research Ethics Committee (2021/ETH00165). We assessed changes in monthly overdose numbers using interrupted time series analysis. We also examined the impact of coronavirus disease 2019 (COVID- 19)- related restrictions, and that of new paracetamol overdose treatment guidelines that could have increased referrals to NSWPIC (published online in December 2019 3 ), using changepoint analysis (details: Supporting Information). A total of 1715 exposures to MR paracetamol were reported to NSWPIC


Data source and data extraction
New South Wales Poisons Information Centre (NSWPIC) is the largest Poisons Information Centre in Australia, receiving approximately 50% of the national 220,000 annual poisoning calls from health care professionals and members of the public 1 .Approximately 65% of NSWPIC calls originate from NSW.We analysed data from the NSWPIC database related to poisonings with over-the-counter analgesics.All single-ingredient paracetamol poisonings (MR and immediate release) are coded under the single "paracetamol" substance code; to identify MR paracetamol poisonings we conducted a free text search of the product and dose fields for *665*, *MR*, *SR*, *CR*, *XR*, *osteo*, *sustain*, *modified*.Records returned were reviewed manually and re-coded.We also extracted calls for other analgesics available without prescription, including ibuprofen, diclofenac, mefenamic acid, naproxen, aspirin, and paracetamol/ibuprofen combinations.Intentional poisoning calls were defined as calls coded as "deliberate self-poisoning" or "intentional: other".Only exposure calls were counted (that is, re-calls and subsequent calls about the same event were excluded).

Interrupted time series analysis
We used interrupted time series analysis to assess changes in monthly poisoning calls following the up-scheduling of MR paracetamol from Schedule 2 to Schedule 3 in June 2020.Interrupted time series analysis is one of the strongest quasi-experimental designs for evaluating the effectiveness of population-level health interventions commenced at a defined time point. 2Our outcome variable, the number of poisoning calls, follows a Poisson distribution.However, when the expected count number is large and the distribution is not bounded by zero, a Poisson distribution can be approximated by a normal distribution.We therefore applied segmented linear regression, which assumes a continuous outcome, to model the intervention.
We used a combination of the Durbin-Watson test, the Ljung-Box test for white noise, and the autocorrelation function and partial autocorrelation function plots to test for the presence of auto-correlation in our models.If any of these tests indicated auto-correlation, we included autoregressive terms in the base model to control for it.We estimated the models using the arima function in the stats R-package.We specified the model order as (p,0,0), with p representing the autoregressive order.We investigated for seasonality by including dummy variables for the months and Fourier terms in our models.We checked the normality assumption and the homoscedasticity assumption of residuals underlying linear regression by inspecting the residual plots against time and against the fitted values, and the normal quantile-quantile plot of residuals.All the analyses were performed in R 4.0.1.

Structural changepoint analysis
Because the up-scheduling of MR paracetamol was implemented shortly after the start of the COVID-19 pandemic and related lockdowns started in NSW, we conducted structural changepoint analysis to identify whether changes in the monthly poisoning call numbers may have preceded up-scheduling, using multiple structural changepoint modelling 3 and the breakpoints function in the strucchange R-package.We fitted time series models adjusted for auto-correlation as necessary and seasonality according to the two approaches described above, and chose the most parsimonious model based on Akaike information criterion (AIC) and after checking other model assumptions (normality and homoscedasticity).We applied multiple structural change modelling to the model chosen by AIC and computed the optimal break points and their 95% confidence intervals (CIs).The changepoint analysis estimates the breakpoint; its variance, a measure of the uncertainty in the estimate, provides the 95% CI.The null hypothesis was no change or break in the series.We allowed for the number of breakpoints to be more than one.Finally, we performed interrupted time series analysis as described above with the identified optimal breakpoint as the intervention date.For each outcome, only one breakpoint was identified.There was a decrease in proportion of poisonings which involved MR paracetamol following re-scheduling: MR paracetamol accounted for 6.9% of OTC analgesic poisonings pre-intervention, and 6.2% post-intervention (p= 0.02, Pearson's chi-squared).There was a slight reduction in median dose taken associated with the re-scheduling, from 13.3 g (IQR 7.3, 23.9 g) to 11.8 g (IQR 6.7, 22.3 g), p=0.008 (Mann-Whitney) (figure 1).

Figure .
Figure.Violin plot showing dose taken in modified release paracetamol overdoses, before and after re-scheduling.Dashed line shows median and dotted lines show upper and lower quartiles.