Estimation of adherence to urate‐lowering therapy in people living with gout using Australia's Pharmaceutical Benefits Scheme and patient‐reported dosing

The aim of this study was to estimate adherence to urate‐lowering therapy (ULT), predominately allopurinol, from Australia's Pharmaceutical Benefits Scheme (PBS) claims database in association with (1) patient‐reported doses and (2) World Health Organization's (WHO) defined daily doses (DDD), namely, allopurinol (400 mg/day) or febuxostat (80 mg/day).


| INTRODUCTION
Gout is characterized by urate crystal deposition, triggering an inflammatory response and painful joint flares. 1 Central to the long-term treatment of gout is the lowering of serum urate (SU) concentration below a target of 0.36 mmol/L (6 mg/dL) achieved using uratelowering therapies (ULTs, e.g., allopurinol or febuxostat) enabling the dissolution of SU crystals, reduction in flare frequency and resolution of gouty tophi. 2,3If gout is left untreated or inadequately treated, the result can be an increased frequency and severity of gout flares, and accumulation of urate deposits, 4 which can damage joints 5 and eventually other organs such as kidneys. 6Despite the 'treat-to-target' SU approach to gout management using ULTs being widely endorsed, implementation remains poor in all countries where it has been examined. 7,8dication adherence can be defined as 'the process by which patients take their medications as prescribed'. 9Suboptimal medication adherence is highly prevalent in chronic conditions and often has detrimental consequences on disease progression, mortality and health care costs. 10,11There have been several methods proposed to measure medication adherence including patient self-reporting, the measurement of plasma drug concentration or drug response biomarkers, electronic monitoring (such as Medication Event Monitoring Systems [MEMS ® ]), 12 pill counts and the retrospective analysis of data from prescription refill records and claims databases. 10,12,13e proportion of days covered (PDC) is commonly employed to measure adherence from analysis of dispensing claims databases. 14PDC is the proportion of days a medication at the prescribed dose is on hand (available) over a defined time interval. 15To calculate PDC over a chosen period, a longitudinal record of each person's dispensings, the number of units of medicine dispensed (e.g.200 tablets) and the strength of the dispensed medicine (e.g. 100 mg) is needed.Also needed is the daily dose during the period, which may be obtained from a medical record, patient survey or extrapolated from the dispensing record itself using knowledge of common dose ranges and/or assumptions such as a defined daily dose (DDD). 16C can be calculated using either the daily dose or WHO's DDD. 17,18The DDD is defined as the assumed average maintenance dose per day for a drug used for its main indication in adults. 16The PDC estimation uses each dispensing to construct an array of days that the medication is available at the daily dose over a chosen time interval.If the dispensing that starts the time interval selected to calculate the PDC is 'overlapped' with the availability of medicines from the previous dispensing, this additional amount is accounted for. 15,19 Australia, prescription pharmaceutical claims data are recorded in the Pharmaceutical Benefits Scheme (PBS) database for administrative purposes.PBS provides subsidized prescription medicines, including allopurinol and febuxostat, for all Australian residents, who hold a Medicare card.PBS-subsidized medicines are dispensed predominantly in community pharmacies (primary care) and private hospitals, and are also dispensed to outpatients and discharging patients (but not to inpatients) in public hospitals in most states and territories. 20Private prescriptions for medicines or indications not listed under the PBS are not covered under this scheme.
[22] Switching medications and dose changes further confound estimates of adherence determined using the PBS claims dataset.In instances where the dose intended to be taken daily is unknown, previous studies have used either guideline-specified (standard) daily doses 23 or the WHO's DDD for that drug and condition. 24[27] Caution is needed as standard dose or DDD may not accurately represent an individual level dosing regimen. 28While the DDD can be adequate depending on the type of drug studied, it is less useful for What is already known about this subject • Using longitudinal dispensing records allows estimation of PDC if prescribed daily dose is known.
• Person-sourced reporting of daily doses can substitute for prescriber's record of the daily dose but has not been combined with longitudinal dispensing records in estimating adherence to ULT.

What this study adds
• This study demonstrated how to estimate adherence to ULT from person-sourced data on dosing and data from longitudinal dispensing records (e.g. from Pharmaceutical Benefits Scheme [PBS] databases in Australia).
• Using self-reported ULT dosing data in combination with longitudinal dispensing data increased confidence in the estimation of PDC in most participants.medicines with wide therapeutic dose ranges. 17The approach of using dispensing claims along with guideline-specified daily doses or WHO DDDs to estimate ULT adherence in gout patients has not been explored.However, ULTs prescribed for gout do not have fixed 'standard' doses.Indeed, guidelines recommend dose escalation starting at allopurinol doses of 50-100 mg/day and titrating up to the dose that achieves the target SU, most commonly 300 mg/day or less 2,29,30 and hence guideline or DDD-based estimates may be inaccurate.
This study aimed to estimate adherence to ULT using PBS dispensing data supplemented with participant-reported daily doses captured in the Gout APP (GAPP) trial's baseline survey.Among participants taking allopurinol, we compared this approach to the one using the WHO DDD for allopurinol as the daily dose.The agreement between patients' ULT-taking status (from self-reports) and ULT-dispensed status (from PBS longitudinal records) was also examined.

| Study design
Data for this study were obtained from the GAPP trial, a randomized controlled trial comparing a test and control app; the test app was designed to improve self-management of gout. 31Entry criteria for GAPP trial participants were a diagnosis of gout, a gout flare in the last 12 months, currently taking or eligible to start or restart ULT, and access to a smartphone device and the internet.Participants in the GAPP trial were given the option to consent for the release of their PBS data over a 3-year period that encompassed their participation in the GAPP trial.Services Australia supplied the PBS data which included all dispensings for each participant in this 3-year period with the dates of dispensings, formulation strength and number of tablets supplied.Only baseline data from the GAPP trial were used for the current study.
The GAPP baseline survey asked participants to record their daily dose of ULT if they reported taking ULT at the time of the survey.Specifically, participants were asked: 'Please select all of the medications you are currently taking for your gout, and record the strength, dose, and frequency'.Choices listed were 'Allopurinol (e.g., Zyloprim, Progout, Allosig)', 'Colchicine (e.g., Colgout, Lengout)', 'Febuxostat (e.g., Adenuric)' or 'None of the above' (Figure S1).After completing their baseline surveys, participants were randomly given access to either the test or control app.

| Ethical considerations
The GAPP trial received approval by the University of New South Wales' Human Research Ethics Committee (HC15199 and HC210543).
All participants provided written informed consent.

| Estimation of PDC
We estimated PDC in GAPP participants in the baseline period of 12 months prior to the app delivery and included participants with PBS data available for at least 11.5 months prior to the app delivery.PDC estimations for periods less than 12 months are not recommended as they provide less accurate estimates. 14The first to last ULT dispensings within this period were used to calculate PDC using the following formula: total days covered from first to last dispensing (excluding the supply from the last dispensing) at the estimated daily dose divided by the total days in the period (first to last dispensing), multiplied by 100 (Figure 1).A 'lookback window' was employed to account for medications on hand from the dispensing immediately prior to the period of interest.Participants with PDC values ≥80% were deemed to be adherent.
For the primary analyses of the current study, we identified people with at least two ULT dispensings in the 12 months prior to app distribution.We estimated PDC for participants who reported F I G U R E 1 Gout APP trial timeline depicting allopurinol dispensings in determining proportion of days covered (PDC).
taking and were dispensed allopurinol and then for those who reported taking and were dispensed any ULT drug (allopurinol or febuxostat).In secondary analyses, we also estimated PDC for participants who had at least three ULT dispensings in the 12 months prior.
As some participants had only one or no dispensings prior to app distribution and some had less than 12 months of any PBS data prior to app distribution, several exploratory analyses on these groups of participants were undertaken, as described in Table S1.

| Estimation of daily dose of ULT
We estimated the most likely ULT dose prescribed per day over the PDC estimation period using the participant's self-reported dose from their baseline survey combined with their PBS dispensing records.
The confidence we had in the dose estimate for each participant was graded as high, moderate, low or none (Figure S2; Table S2).If selfreported dosing data was complete, plausible and matched their PBS dispensing data, we rated our confidence in the daily dose estimate to be high (Table S3).If the tablet strength reported was not available in Australia (e.g.600 mg of allopurinol), we verified if the selfreported daily dose was plausible given the dispensed tablet size (e.g.600 mg/day by taking 2 Â 300 mg tablets/day).Consensus decisions were made and documented by the investigators (MS, RD, EA, MC) for each participant with clinical input from the rheumatologist investigator (RD).The same process was undertaken for the sensitivity analyses.
We estimated PDC using the self-reported dosing data as described above (Method 1) and, for comparison, calculated PDC using the WHO's DDD for participants taking allopurinol (400 mg) or febuxostat (80 mg) (Method 2).We used a paired t-test and Wilcoxon signed rank test to examine the difference in mean and median PDCs between the two methods in participants who reported taking and were dispensed allopurinol, and then in participants who reported taking and were dispensed allopurinol or febuxostat.
The level of agreement between participants' ULT-taking status (from self-reports) and ULT-dispensed status (from PBS records) in a 12-month period was examined by calculation of positive percent agreement and negative percent agreement, 32 as well as testing association with Cohen's kappa (κ).Pearson's chi-square test was used to test whether participants who reported taking more than 300 mg/day of allopurinol in the baseline survey were more likely to be adherent (PDC ≥80%) or more likely to have an SU concentration below 0.36 mmol/L at baseline.

| Selection of participants for estimation of PDC
Among 537 GAPP participants, 133 reported taking ULT and had at least two dispensings of ULT in the 12 months prior to sending them an app (Figure 2).A review of their PBS dispensing records showed that 20 of these participants had a dosage change (i.e.change in tablet strength dispensed) (n = 16), or coincident ULT dispensings (i.e.all ULT dispensings occurred on the same day) (n = 3), or a ULT switch (n = 1) leading to less than two dispensings.Thus, this group of participants (n = 20) now had less than two dispensings and were no longer eligible for estimation of PDC.Five more participants were excluded as their daily dose could not be established with any confidence ('None' confidence classification in Table S2), leaving 108 eligible for PDC calculation (Figures 2 and 3).

| Patterns of ULT dispensings
Of 402 GAPP participants who were dispensed any ULT in their PBS data extraction period, 389 were dispensed only allopurinol.Most of these participants were only dispensed 100 mg tablets of allopurinol (n = 271), a quarter were only dispensed 300 mg tablets (n = 99), while the remainder were dispensed a combination of 100 mg and 300 mg tablets (n = 19).

| Estimation of PDC in 108 eligible participants
In the two or more dispensings group (n = 108), the time span for estimation of PDC was a mean (SD) of 206 (92) days (range = 19-349 days).Only 10 (9.3%) in this group needed a 'lookback' adjustment to account for an overlap from the last dispensing prior to the start of the 12-month PDC estimation period. 14For the three or more dispensings group (n = 77), which was a subset of the two or more dispensings group, the time span was a mean (SD) of 241 (72) days (range = 64-349 days).The mean number (SD) of dispensings was 4.5 (2.4) in the two or more dispensings group, and 5.3 (2.3) in the three or more dispensings group.

| ULT-taking status (self-report) versus ULT-dispensed status (PBS data)
Of the 458 participants who were sent an app, 355 had at least a 12-month window prior to the app being sent where any PBS dispensing that had occurred would be available for review.Of these, F I G U R E 2 Flow diagram depicting the data available to determine proportions of days covered at baseline of Gout APP trial.Abbreviations: DD, daily dose; DDD, defined daily dose; MBS, Medicare benefits schedule; PBS, Pharmaceutical Benefits Scheme; ULT, urate-lowering therapy.135 (38%) were not dispensed any ULT in these 12 months.The positive percent agreement and negative percent agreement of selfreported ULT-taking status against ULT-dispensed status in the 12 months prior to the app being sent was 88.9% and 81.7%, respectively (Table 2).There was also good agreement 33 between selfreported data and ULT dispensings (Cohen's κ = 0.708, P < .001).

| Confidence in daily dose estimation
Out of the 108 participants whose PDC was calculated using two or more dispensings in the 12 months before the app was sent, 27 (25%) had daily dose estimates that were given confidence classifications of either moderate (n = 19) or low (n = 8; Table S2).Similar estimates of PDC were observed when only low confidence cases or both low and moderate confidence cases were excluded from the PDC estimation (Table 3).

| Exploratory analyses
Further exploratory analyses in participants who had only one or no dispensings prior to app distribution and those with less than 12 months of any PBS data before the app distribution resulted in similarly high PDC estimates (Table S1).Participants with a daily dose above 300 mg were not more likely to be at or above target SU concentrations (Table S5).In addition, we found participants deemed adherent to ULT were also not more likely to be at or below target SU (Table S6).

| DISCUSSION
To our knowledge, this is the first study to estimate ULT adherence using self-reported daily dose and pharmaceutical claims data in a sample of people living with gout.Understanding the prevalence of, and trends in, adherence to ULT is an important tool in the effort to improve rates of adherence at a population level and thereby outcomes for people suffering with gout.
Participants taking allopurinol in our cohort had a mean PDC of 83% estimated using self-reported daily dose.This fell to 63% using the WHO DDD of 400 mg/day.Using Method 1 (self-report), 63% of our cohort were adherent (PDC ≥80%), while only 23% were adherent when the WHO DDD was used in the calculation (Method 2).[36] The results suggest strongly that the WHO DDD estimate exceeds actual doses employed to manage gout.
Lee and So 34 found that among 132 Korean men with gout, 94 (71.2%) were adherent to ULT (medication possession ratio [MPR] ≥80%) 1 year after treatment initiation.MPR is calculated similarly to PDC but not adjusted for medication dispensed prior to the start date for calculation of MPR, thereby tending to overestimate adherence in comparison to PDC. 13 Two other studies reported mean MPRs of 96% 36 and 88%, 35 with 84% 36 and 78% 37 of the study participants found to be adherent to ULT using the MPR ≥80% standard.However, recent systematic reviews and meta-analyses based on prescription claims data [38][39][40] reported much lower proportions of people classified as ULT-adherent (ranging from 10% to 48%).In the most recent systematic review, 39 the mean ULT adherence rates from seven studies (using MPR or PDC estimates) varied from 54% to 88%.
These differences between studies are likely related to the characteristics of the cohorts examined.Adherence rates observed in this study are likely to be better than in clinical practice.Our participants were volunteers in a research study testing an app designed to improve self-management of gout, so were more likely to be adherent than a random sample of people with gout in the community.
Indeed, the majority (65.4%) of our sample reported being currently employed or students, and 72.9% reported tertiary education as their highest level of education completed (Table 1).Furthermore, our study participants primarily reported White/Caucasian/European ancestry as their sole ancestry.We recommend using the methods we describe to determine adherence to ULT in groups with a high prevalence of gout, including Aboriginal and/or Torres Strait Islander and M aori and Pacific Islander peoples living in Australia. 35,41,42lf-reported daily doses of ULT drugs in the current study were comparable to other recent studies. 30,43,44The overall self-reported mean (range) allopurinol dose in our study was 291 mg/day (50-1200 mg/day) and for febuxostat 80 mg/day (40-120 mg/day).The majority (85%) of allopurinol takers reported a dose below the WHO DDD of 400 mg/day.While these constituted almost all participants who were classified as adherent (95%) using Method 1, only 26 (43%) reached target urate.
Our dose findings also reflect those reported by Stamp and colleagues. 43The mean (range) allopurinol dose of 183 trial participants in New Zealand was 269 mg/day (100-600 mg/day).Among all participants in the GAPP trial who dispensed any ULT in their PBS data extraction period (n = 402), two-thirds were dispensed only 100 mg tablets of allopurinol, a quarter dispensed only 300 mg tablets, and 0.05% dispensed a combination of 100 and 300 mg tablets.This may reflect the growing prescription of lower allopurinol doses in Australia, a trend observed by two Australian drug utilization studies. 30,44An important question is whether the dose of ULT is sufficient to reach target SU concentrations.The higher the dose, the more likely target SU will be achieved.In our baseline data for the GAPP trial, those with a daily dose estimated to be above 300 mg/day were not more likely to have SU at or below target concentrations (Table S5) and similarly those with PDC ≥80% were not more likely to be at target SU (Table S6).This result is likely reflective of the small numbers of participants taking more than 300 mg allopurinol/day.
One limitation was that we assumed self-reported daily doses from the baseline survey reflected daily doses of medications dispensed in the approximate 12-month period before the app was sent.
We endeavoured to confirm self-reported daily doses of allopurinol by comparing these reports (e.g.300 mg/day) with dispensed pack sizes and timing of dispensings revealed in the PBS data.As presented in Table 2, we found that there was a good agreement between selfreported ULT-taking status and ULT-dispensed status within the 12 months prior to the app being sent (κ = 0.708, P < .001;positive percent agreement = 88.9%,negative percent agreement = 81.7%).
Our results also appear to corroborate the findings of Gnjidic and colleagues. 45 The format of our baseline survey may have had an impact on the accuracy of self-reported daily dose.Participants had the option to enter free text including words and/or numbers into the text boxes (Figure S1) that led to some uncertainty about interpretation of doses in some cases.We sought to correct obvious errors in survey responses by contacting participants.The research team rated the accuracy of the self-reported dose estimate to be low in only eight out of 108 participants (Table S3).Removal of these eight had virtually no impact on PDC estimates in sensitivity testing.
Since the PBS datasets do not include the prescribed daily dose, researchers utilizing PBS data for adherence calculations need to use a proxy.In Australia, it is common practice to use the standard daily dose derived from recommendations provided by the Australia Medicines Handbook or therapeutic guidelines. 23Alternatively, WHO DDDs can be used.These DDDs were developed by the WHO for comparing drug use between populations in drug utilization studies.DDD are defined as 'the assumed average maintenance dose per day for a drug used for its main indication in adults'. 46However, DDDs may not be indicative of the therapeutic or recommended daily dose.DDDs are considered general and at times inaccurate estimates of usual doses. 47e DDD for allopurinol (400 mg/day) reported by the WHO 48 is of limited use as no standard daily dose exists for this ULT; optimally, and according to the guidelines, the dose of allopurinol should be titrated up until target SU concentration is achieved.More indicative estimates of daily doses of allopurinol were reported in a recent longitudinal analysis of national prescribing patterns by Coleshill and colleagues 30 but this study did not report effectiveness in reaching target SU.
Our study demonstrated that self-reported ULT dosing data is a helpful adjunct for estimating ULT adherence especially for medicines that do not have a recommended fixed daily dose.Our work, however, highlights the challenges inherent in using PBS claims data, notably uncertainty about the daily dose prescribed and taken when using dispensing claims alone.Dispensing data sets that recorded at least the prescribed daily dose would be a significant improvement on our ability to estimate adherence.In mitigating this deficiency, patientreported dosing helped increase confidence in the daily dosing estimates; however, this approach is also limited as it relies on crosssectional survey to estimate the daily dose over a significant time period.Future research should focus on longer time intervals to calculate PDCs for ULT given the very large pack sizes of at least the 100 mg formulation, improved survey instruments, and other innovative approaches to obtaining the prescribed and/or taken daily doses (e.g. using electronic medication diaries).
A systematic review from 2018 reported that people who are older, have a greater number of comorbidities and suffered from diabetes or hypertension were more likely to be adherent to ULT. 39 However, no Australian studies met the criteria for inclusion in this review.In the Australian setting, it has been reported that a lack of education about gout and its treatment was a barrier to care in gout, 49 and ULT use was associated with male gender, low socioeconomic status and the concomitant use of cholesterol-lowering therapy. 50Factors affecting ULT adherence in Australia need to be further explored.
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20.

F I G U R E 3
Decision processes for determining PDC values.Abbreviations: DD, daily dose; PBS, Pharmaceutical Benefits Scheme; PDC, proportion of days covered; ULT, urate-lowering therapy.T A B L E 1 Sociodemographic and clinical characteristics of participants by ULT (allopurinol [n = 101] and febuxostat [n = 7]) dispensing group.Patient factors Two or more ULT dispensings in the 12 month interval (n = 108)

a
One participant did not complete this question in the baseline survey from the two or more ULT dispensings group.b One participant did not complete this question in the baseline survey from the two or more and the three or more ULT dispensings group.
PDC values for participants taking allopurinol, including and excluding estimates of daily dose given a 'low' confidence rating.