A systematic scoping review of interventions to optimise medication prescribing and adherence in older adults with cancer

4 Background: Older adults with cancer often require multiple medications (polypharmacy) 5 comprising cancer-specific treatments, supportive care medications (e.g. analgesics), and 6 medications for pre-existing health conditions. Increasing numbers of medications may increase 7 risks of potentially inappropriate prescribing and non-adherence. 8 Objective(s): To provide an overview of evaluations of interventions aimed at optimising 9 medication prescribing and/or adherence in older adults with cancer. 10 Methods: A systematic scoping review was undertaken. Four databases (PubMed, EMBASE, 11 CINAHL, PsycINFO) were searched using relevant search terms (e.g. cancer, older adults). 12 Eligible studies evaluated interventions seeking to improve medication prescribing and/or 13 adherence in older adults (≥65 years) with cancer using a comparative evaluation. All outcomes 14 for studies that met inclusion criteria were included in the review. Extracted data were collated 15 using tables and accompanying narrative descriptive summaries. The review was reported using 16 the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for scoping 17 reviews (PRISMA-ScR) guidelines. 18 Results: Nine studies met inclusion criteria comprising five randomised controlled trials (RCTs) 19 and four before-and-after study designs. Studies were primarily conducted in oncology clinics, 20 ranging from single study sites to 109 oncology clinics. Sample sizes ranged between 33-4844 21 patients. Interventions most commonly involved patient education (n=6) delivered by 22 pharmacists or nurses. Three studies reported on prescribing-related outcomes and seven 23 studies reported on adherence-related outcomes, using different terminology and assessment 24 methods. Prescribing-related outcomes focused on medication appropriateness (using Beers 25 criteria) and drug-related problems including drug interactions. Adherence-related outcomes 26 included assessments of self-reported medication adherence and calculation of patients’ 27 medication possession ratio. 28 Jo urn al Pr e-p roo f


48
The global burden of cancer is rapidly increasing and driven to a large extent by advancing 49 population age (1). In 2012, 48% of cancer diagnoses worldwide occurred in older adults (≥65 50 years), and this is expected to rise to 58% by 2038 (2). Older adults with cancer often have 51 existing health conditions, necessitating the use of multiple medications (3-5). Treatment 52 burden (i.e. the workload of healthcare and its impact on patient functioning and well-being) 53 (6) in this patient cohort can be further potentiated by the addition of cancer-specific 54 treatments and supportive care medications (e.g. analgesics). A sizeable proportion (21%) of 55 admissions to cancer services are related to adverse drug events arising from the use of both 56 cancer-specific and non-cancer treatments (7). Therefore, interventions aimed at optimising 57 medication regimens in older patients with cancer are required. 58 Medicines optimisation is defined as an approach to care that focuses on ensuring the best 59 clinical outcomes for patients through safe and effective medication use (8). Two key aspects of 60 medicines optimisation involve reducing potentially inappropriate prescribing and enhancing 61 medication adherence (9). Potentially inappropriate prescribing (PIP) encompasses a range of 62 suboptimal prescribing practices, including prescribing medication associated with high risk of 63 adverse drug events, inappropriate doses or treatment durations and errors of omission 64 (under-prescribing) (10). Several criteria have been developed to assess the appropriateness of 65 medication prescribing in the general older population (11)(12)(13)(14), and application of these criteria 66 has proven to be effective in reducing PIP (15). However, many of these criteria were not 67 designed specifically for use in older adults with cancer, and therefore, their applicability to this 68 patient cohort is limited. Prescribing for older adults with cancer requires additional 69 considerations such as the risk of drugs becoming inappropriate as a consequence of 70 physiological changes (e.g. weight loss) due to cancer and the associated treatments (16), as 71 well as the increased potential for drug-drug interactions arising from the use of systemic 72 anticancer therapies and supportive treatments (17). Furthermore, existing tools often 73 recommend the avoidance or cessation of certain medications that may be clinically suitable for 74 this patient cohort (e.g. non-steroidal anti-inflammatory drugs for cancer-related pain) (18,19). 75 The OncPal deprescribing guideline has been developed which is specific to cancer populations 76 J o u r n a l P r e -p r o o f methodological guidance available from the Joanna Briggs Institute (28, 29) and is reported 104 using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for 105 scoping reviews (PRISMA-ScR) guidelines (30). 106 Search strategy 107 Four electronic databases (PubMed, EMBASE, CINAHL and PsycInfo) were searched from 108 inception to 29 th November 2019 using established search methods for scoping reviews (28). 109 Preliminary searches of each database were undertaken to identify relevant keywords and 110 index terms. Initial key search terms included: "cancer", "older adults", "prescribing" and 111 The review included interventions targeting older adults (≥65 years) with an active cancer 117 diagnosis. Eligible studies were not limited by type of cancer, healthcare setting or 118 number/types of medication prescribed. In order to meet inclusion criteria, studies had to 119 include individuals aged ≥65 years or a study population with a mean/median age of ≥65 years. 120

Types of interventions 121
The review included any interventions seeking to optimise medication prescribing or adherence 122 in older patients with cancer (or a study population with a mean/median age of ≥65 years). To 123 be included, interventions had to align with key principles of medicines optimisation which 124 were broadly categorised as: interventions to reduce PIP and interventions to improve 125 medication adherence (8). Interventions could target either the level of the healthcare 126 professional or patient. 127 J o u r n a l P r e -p r o o f

Types of studies 128
Studies had to include some form of comparative evaluation (e.g. inclusion of a control group or 129 use of a controlled before/after design). Only studies published in the English language were 130 eligible for inclusion. Published conference abstracts were not included. 131

Types of outcomes 132
As there is no existing overview of outcomes of interventions aimed at improving prescribing or 133 adherence in older patients with cancer, all outcomes for studies that met the above inclusion 134 criteria were included in the review. This allowed for an overview to be provided of the range of 135 outcomes that have been evaluated on this topic. 136

Study selection 137
Following deduplication, all abstract titles were screened initially by one review author (X) to 138 remove any that were clearly irrelevant. Of the remaining abstracts, a randomly selected 50% 139 sample were double screened by two review authors working independently (Y, Z). If an 140 abstract appeared to meet inclusion criteria or could not be excluded based on the title or 141 abstract alone, the full-text article was retrieved and independently assessed for inclusion by 142 two reviewers. Any disagreements throughout this process were resolved through consensus 143 discussion with other members of the research team. 144 Charting, collating and summarising the data 145 Two reviewers (X, Y) independently performed data extraction (often referred to as "charting" 146 in the scoping review literature (28)), using a data extraction form that was developed in 147 accordance with relevant methodological guidance (31). Data were extracted relating to each of 148 the following: authors; year of publication; country of origin; study design and setting; study 149 aims/purpose; study population; sample size; outcome measures; intervention characteristics; 150 key findings. Interventions were categorised according to the Effective Practice and 151 Organisation of Care (EPOC) taxonomy (i.e. 'Delivery arrangements', 'Financial arrangements', 152 'Governance arrangements', 'Implementation strategies') (32). 153 Formal assessments of methodological quality of included studies were not undertaken, as the 154 aim of a scoping review is to provide a broad overview of the existing literature (28). A narrative 155 synthesis was completed following established guidance (33) which involved three key steps 156 (detailed further in the published protocol (27)): (i) development of a preliminary synthesis of

160
Search results

161
The electronic searches yielded 21,136 citations ( Figure 1). Following title and abstract 162 screening, 309 full-text articles were reviewed for eligibility. In total, nine studies met inclusion 163 criteria (34-42). The main reasons for exclusion of the remaining articles were: study population 164 was not cancer specific and/or aged 65 years; intervention did not aim to optimise prescribing 165 or adherence; study design did not include a comparative evaluation. 166 [insert Figure 1] 167  Key study findings relating to the review question are summarised in Table 3  DRPs were drug-drug interactions (57.3%), dosage errors (11.2%) and administration errors 281 (11.2%). Of the identified DRPs, 74% were rated as significant and 24% were rated as life-282 threatening. 283 All four of the non-randomised studies reported on adherence-related outcomes using 284 comparisons between intervention groups and historical control groups. One study reported 285 that a higher proportion of intervention group patients were adherent compared to the control 286 group, at six-months post intervention (95% vs 87.7% respectively, p=0.025) (38). Another study 287 reported no difference in therapy persistence between intervention and control groups at the 288 three-month follow-up (73% vs 59% respectively, p=0.7) (34). The other two studies reported 289 no differences in post-intervention adherence to pazopanib (40) or adherence to calcium and 290 vitamin D intake at three-month follow-up (42). 291 [insert Table 3] 292 293 J o u r n a l P r e -p r o o f  The intervention was delivered using a combination of faceto-face education and written material by an oncology nurse or physician who were part of the oncology team.  Table 3: Overview of key findings from included studies 00

Study ID Key findings relating to medication prescribing and adherence Conliffe 2019 (34)
Adherence-related outcomes Persistence on therapy: At three month follow-up, a higher proportion of intervention group patients remained on the prescribed oral antineoplastic medication compared to the control group (73% vs 59% respectively, p=0.7).
Ferrell 1993 (35) Adherence-related outcomes Medication compliance: Preliminary results indicated that many patients were only taking 70% of the medications prescribed. Intervention and control group data were not reported separately.
Hadji 2013 (36) Adherence-related outcomes Compliance: At 12-month follow-up, overall compliance across both groups was 88.6%. There was no difference in compliance between the intervention and control group (88.5% vs 88.8% respectively, p=0.81) Persistence: At 12-month follow-up, overall persistence across both groups was 41.8%. There was no difference in persistence on therapy between the intervention and the control group (40.5% vs 43% respectively, p=0.18).

Nipp 2019 (37) Prescribing-related outcomes
Potentially inappropriate medications: At four-week follow-up, intervention patients had fewer potentially inappropriate medications than control group patients (3.46 vs. 4.80, p = 0.069), although differences were not statistically significant.
Discrepant medication: At four-week follow-up, intervention patients had fewer discrepant medications than control group patients (5.82 vs. 8.07, p = 0.094), although these differences were not statistically significant.

Ribed 2015 (38) Prescribing-related outcomes
Drug-related problems: Over the six-month follow-up period, a significantly higher number of drug-related problems were detected in the intervention group (n=169) compared to the control group (n=106, p=0.008). Drug-related problems primarily consisted of drug-drug interactions (57.4%), incorrect administration or time of use (11.2%), and Riese 2017 (39) Adherence-related outcomes Unplanned therapy interruption: The intervention group had a lower chance of experiencing a patient-initiated unplanned therapy interruption two weeks' post-intervention (Odds ratio; 0.14, 95% confidence interval; 0.03, 0.69; p=0.01). This was not sustained at three follow-up assessments over a three-month period whereby no significant differences were observed between intervention and control groups (p >0.05).
Todo 2019 (40) Adherence-related outcomes Adherence: There was no reported post-intervention change in adherence in the intervention group, whereas nonadherence occurred in five out of 13 patients (38%) in the control group.
Trowbridge 1997 (41) Prescribing-related outcomes Prescribing patterns of analgesics: There was a significant post-intervention difference in analgesic prescribing (p=0.0162) with changes in prescriptions observed in 25% of intervention group patients (reduction in prescribing in 5% and an increase in prescribing in 20%) compared to 14% of control group patients (all involved increases in prescribing).
Under-treatment of pain: No significant post-intervention difference in the under-treatment of pain between intervention and control group (35% vs. 38% respectively) Tsang 2018 (42) Adherence-related outcomes Self-reported adherence to vitamin D intake ≥ 1000 IU/day: At three-month follow-up, the intervention had no Intervention components included patient education and medication reviews and were 332 primarily delivered by pharmacists and nurses. However, in many cases, detailed descriptions of 333 intervention were lacking. It has yet to be determined which healthcare professionals would be 334 best suited to delivering interventions to this group of patients and how frequently 335 interventions should be delivered. Based on previous research conducted among the general 336 older population (i.e. non-cancer specific), it has been suggested that multifaceted 337 interventions comprising several intervention components may be more likely to improve 338 prescribing and adherence than single component interventions (10,54). In order to enhance 339 adherence to oral cancer therapy in older adults, the International Society of Geriatric Oncology 340 has recommended that self-management interventions (e.g. patient education, self-monitoring 341 for toxicity) be combined with adherence interventions led by healthcare professionals (55) . It 342 also recommended that technologies such as text messaging, electronic mail and other 343 automated alert systems be used to promote adherence within this population. The 344 interventions identified in this review were most commonly delivered either face-to-face or by 345 telephone and none of them reported including a technological component. A recent review 346 examining the use of information technology in the monitoring of older patients with cancer 347 concluded that technology could be beneficial in this patient cohort, but any technology 348 employed needs to be tailored to the unique needs of this population (e.g. text should be of 349 appropriately large font size, audio prompts with adjustable volume) (56). 350 One particular challenge that arose in synthesising information across included studies related 351 to differences in terminology. For example, in terms of medication adherence, different 352 parameters were examined across studies including compliance (now superseded by 353 adherence), persistence and time-to-treatment discontinuation. Medication compliance and 354 persistence are two different constructs (57) whereby compliance refers to 'the extent to which 355 patients act in accordance with the prescribed dosing regimen instructions', and persistence 356 refers to 'the act of continuing the prescribed treatment for the intended duration' (57). In 357 addition to this, most studies relied on self-reported data, with only one study using an 358 objective measure of adherence in the form of the medication possession ratio (38). Although 359 there is no gold standard assessment for adherence, it is recommended that a combination of 360 measures is used to overcome the limitations of any single assessment tool (58). 361 Variation also existed in the focus of prescribing-related outcomes and the tools that were used 362 across included studies. For example, one study examined drug-related problems, including 363 drug interactions (38), while another study examined prescribing of analgesics as part of a pain 364 management intervention (41). The only study that assessed the appropriateness of medication 365 prescribing (37) used the Beers criteria (18). The Beers criteria are a well-established and widely 366 used form of explicit prescribing criteria for older adults. However, they were not specifically 367 designed or intended to assess prescribing for older adults with cancer and do not target