Disclosures No author reports conflict of interest.
Many factors are associated with medication non-adherence in Parkinson's disease (PD), including complex treatment regimens, mood disorders and impaired cognition. However, interventions to improve adherence which acknowledge such factors are lacking. A phase II randomised controlled trial was conducted investigating whether Adherence Therapy (AT) improves medication adherence and quality of life (QoL) compared with routine care (RC) in PD.
Eligible PD patients and their spouse/carers were randomised to intervention (RC plus AT) or control (RC alone). Primary outcomes were change in adherence (Morisky Medication Adherence Scale) and QoL (Parkinson's Disease Questionnaire-39) from baseline to week-12 follow up. Secondary outcomes were MDS-UPDRS (part I, II, IV), Beliefs about Medication Questionnaire (BMQ), EuroQol (EQ-5D) and the Caregiving Distress Scale. Blinded data were analysed using logistic and linear regression models based on the intention-to-treat principle.
Seventy-six patients and 46 spouse/carers completed the study (intervention: n = 38 patients, n = 24 spouse/carers). At week-12 AT significantly improved adherence compared with RC (OR 8.2; 95% CI: 2.8, 24.3). Numbers needed to treat (NNT) were 2.2 (CI: 1.6, 3.9). Compared with RC, AT significantly improved PDQ-39 (−9.0 CI: −12.2, −5.8), BMQ general harm (−1.0 CI: −1.9, −0.2) and MDS-UPDRS part II (−4.8 CI: −8.1, −1.4). No significant interaction was observed between the presence of a spouse/carer and the effect of AT.
Adherence Therapy improved self-reported adherence and QoL in a PD sample. The small NNT suggests AT may be cost-effective. A larger pragmatic trial to test the efficacy and cost-effectiveness of AT by multiple therapists is required.
Non-adherence to medication is prevalent in Parkinson's disease (PD).
Various interventions have been widely investigated in chronic conditions for improving adherence.
Very few studies of interventions to improve medication adherence have focused on PD.
Adherence Therapy (AT) is a novel approach to maximising adherence that has shown benefit in other chronic conditions.
This randomised controlled trial is the first to investigate AT in PD.
Adherence therapy significantly improved both medication adherence and quality of life in people with PD. Specifically, patients who received AT reported improvements in mobility, activities of daily living, emotional wellbeing, cognition, communication and body discomfort.
General beliefs about medication also significantly improved in those who received AT compared with controls.
Symptoms of anxiety may predict poorer adherence to medication after receiving AT.
Parkinson's disease (PD) adversely affects Quality of Life (QoL). Whilst treatment commences as monotherapy, polypharmacy becomes essential in later stages. Although medication adherence is often assumed, non-adherence is prevalent in PD . Patients may omit, alter or discontinue treatments, resulting in poor symptom management [2, 3]. Irregular dosing contributes to motor fluctuations, and dyskinesia may result from overmedicating . Optimising adherence is therefore essential in PD.
Interventions for improving adherence have been widely investigated [5, 6]. Though, few studies have focused on PD. We recently reported factors associated with medication non-adherence in PD . However, interventions to improve adherence that acknowledge such factors are lacking.
Adherence Therapy (AT) is a novel approach to maximising adherence. We incorporated our greater understanding of non-adherence in PD into the principles of AT. Through a phase II randomised controlled trial (RCT), we hypothesised that there would be a statistically significant difference in medication adherence and QoL in PD patients who received AT in addition to routine care (RC) compared with those who received RC alone.
The RCT (ISRCTN07830951)  ran from September 2011 to March 2013 within the Neurology and Older People's Medicine departments at a University Hospital in England.
Patients were posted packs containing an invitation letter, the study information sheet and the Morisky Medication Adherence Scale (MMAS-4) , a self-report tool used previously in PD [10, 11] consisting of four yes/no items (Appendix). Patients were adults with idiopathic PD (UK Brain Bank criteria) , prescribed one or more antiparkinsonian medications, had a stable regimen and did not have dementia. Patients returned the MMAS-4 as an expression of interest.
Eligible patients (MMAS-4 score ≥ 1) were seen by DJD following their routine clinic consultations. Participants signed an informed consent form and completed all baseline data prior to randomisation if they agreed to participate in the trial. Originally a cut-off of MMAS-4 score ≥ 2 was used, as we previously described in the protocol . However, to ensure all eligible patients were included, the Trial Steering Committee (TSC) lowered the score to ≥ 1 in line with the author's validation . Patients who returned their MMAS-4 by post after their consultant appointment date were later visited at home instead of being approached in clinic. If they agreed to participate in the trial, all baseline data were collected during the home visit.
Randomisation and stratification
After completing baseline data, participants were then randomised to intervention (RC plus AT) or control (RC alone). Computer-generated random numbers were accessed via a web-based system developed by the Clinical Research Trials Unit at the University of East Anglia. Allocations were made via permuted blocks of four to six. Participants were stratified into ‘spouse/carer present’ or ‘no spouse/carer’ present.
Ethical approval, monitoring and funding
Cambridge Central Research Ethics Committee granted ethical approval (11/EE/0179). A TSC monitored study progress and adverse events.
Participants randomised into this arm received seven weekly one-to-one AT sessions in their home in addition to RC. Where a spouse/carer participated, both received AT together. DJD delivered AT under supervision of a senior therapist/investigator (RJG). AT is a brief, cognitive-behavioural approach aimed at facilitating a process of shared decision making. Modifying beliefs and exploring ambivalence towards medication are key components. AT incorporates four chief skills; namely, keeping patients engaged/minimising resistance, exchanging information, using Socratic dialogue to generate belief discrepancies and identification/amplification of the personally relevant benefits to treatment . There are four areas of focus:
Assessment: medication review, exploring attitudes towards medication, side effects and agenda setting.
Problem-solving: practical issues leading to non-adherence.
Medication timeline: reflection on experience with medication.
Medication ambivalence/beliefs and concerns: discussing pros & cons to taking/not taking medication.
Participants were defined as completing AT if they received at least five of the seven sessions.
Participants in the control arm received RC according to usual practice in the UK, i.e. patients attended consultant and nurse specialist appointments as scheduled based on their clinical needs. No additional contact or information regarding medication adherence was provided.
The MMAS-4, Parkinson's Disease Questionnaire (PDQ-39), Movement Disorder Society-Unified PD Rating Scale (MDS-UPDRS) (parts I, II, IV) , Beliefs about Medication Questionnaire (BMQ) , EuroQol (EQ-5D) , Montreal Cognitive Assessment Scale (MoCA)  and the Hospital Anxiety and Depression Scale (HADS)  were completed at baseline. Spouse/carers completed the BMQ and the Caregiving Distress Scale (CDS) .
We anticipated that improving adherence may impact on QoL. We therefore assessed change in the MMAS-4 and PDQ-39 from baseline to 12-weeks post-randomisation equally as primary outcomes. Secondary outcomes were change from baseline to 12-weeks post-randomisation in MDS-UPDRS, BMQ, EQ-5D and the CDS. Outcomes were also assessed immediately post-intervention (week-7).
Blinding to group allocation was not possible as participants knew whether or not they were receiving AT. DJD delivered AT and collected self-reported follow-up measurements and so was unmasked to group allocation. A blinded medical statistician analysed the trial data.
Eighty participants provided 81% power (alpha 0.05) to detect a 25% difference in adherence between the groups. This provided 80% power to detect a Cohen's effect size of 0.69, typically considered large, in the PDQ-39 overall score based on the published standard deviation (8.89) in a PD cohort . With this, we expected to observe a difference in means of 6.13 (8.89 × 0.69) units in the PDQ-39.
We interpreted adherence to be binary: MMAS-4 scores either ‘did not change or increased’ or they ‘decreased’ (improved adherence). This was compared between groups using logistic regression, adjusting for the presence of a spouse/carer. Numbers needed to treat (NNT) were estimated. All other outcomes were continuous and were compared using linear regression, also adjusting for the presence of a spouse/carer. Model assumptions were found to be robust by comparing the results to those using a non-parametric bootstrap. Subgroup analyses tested for an interaction between subgroup factors [carer or no carer, baseline HADS (0–7 vs. ≥ 8) and MoCA (0–25 vs. ≥ 26)] and the treatment effect of AT. Analyses were conducted according to intention-to-treat by a blinded statistician (ABC).
In total, 427 patients were invited to participate. Of the 249 (58%) who responded, 173 (69%) reported sound adherence (MMAS = 0). The remaining eligible patients (n = 76) were randomly assigned to intervention (n = 38) or control (n = 38) (Figure 1). No patients withdrew from the trial. Twenty-five (66%) patients randomised to intervention had a spouse/carer compared with 23 (61%) in the control. One carer from each group did not complete the study. No patient received fewer than five therapy sessions [average duration/session = 40 min (range 30–60)]. Sample characteristics and baseline data were comparable between the groups (Table 1).
Table 1. Baseline characteristics and measurements
Active group, n = 38
Control group, n = 38
Means and SD unless stated otherwise. *Movement Disorders Society – Unified Parkinson's Disease Rating Scale (MDS-UPDRS) part IV. †Median and interquartile range (IQR); MoCA, Montreal Cognitive Assessment Scale; HADS, Hospital Anxiety & Depression Scale; LEDD, Levodopa Equivalent Daily Doses; MMAS-4, Morisky Medication Adherence Scale; PDQ-39, Parkinson's disease Questionnaire-39; BMQ, Beliefs about Medication Questionnaire; CDS, Caregiving Distress Scale.
Week-12 analysis showed 60.5% of the intervention group improved in adherence compared with 15.8% of the controls [Odds ratio (OR) 8.2; 95% confidence interval (95% CI) 2.8, 24.3]. The NNT was 2.2 (95% CI: 1.6, 3.9) (Table 2). Week-7 analysis revealed adherence improved in 64.8% of the intervention group compared with 26.3% in the controls (OR 6.1; 95% CI: 2.2, 16.4). The NNT was 2.4 (95% CI: 1.6, 4.6) (Table 3).
Table 2. Outcomes (change from baseline to week-12 follow up)
Week-12 analysis showed the PDQ-39 significantly improved in the intervention group but worsened slightly in the controls (−9.0; 95% CI: −12.2, −5.8). The PDQ-39 also improved at week-7 in the intervention group but worsened slightly in the controls (−4.2; 95% CI: −7.2, −1.3). Analysis of the eight PDQ-39 domains showed patients in the intervention group significantly improved in mobility, activities of daily living, emotional wellbeing, cognition, communication and body discomfort compared with the controls (Table 2).
For secondary outcomes, BMQ (general harm) improved in the intervention group but worsened slightly in controls (−1.0; 95% CI: −1.9, −0.2). There was no significant difference between the groups for BMQ general overuse, specific concern and specific necessity. The MDS-UPDRS part II (motor experiences of daily living) improved in the intervention group but worsened slightly in controls (−4.8; 95% CI: −8.1, −1.4). There was no significant difference between the groups in MDS-UPDRS (part I, IV), EQ-5D or the two spouse/carer outcomes. Although not analysed statistically, at week-12 seven fewer participants in the intervention group reported dyskinesia compared with one extra in the control group. Furthermore, six fewer participants reported motor fluctuations in the intervention group compared with three extra in the control group.
For analyses testing the potential interaction between subgroup factors (carer or no carer, baseline HADS and MoCA) and the treatment effect of AT, only a score less than 7 on the HADS anxiety domain (no anxiety) compared with ≥ 8 (anxiety) predicted greater adherence (OR 21.1; 95% CI: 4.1, 107.6). There was no statistically significant interaction of either subgroup factor on PDQ-39 scores.
Although no health economic analyses were conducted, the cost of therapy (including time/travel) for a PD nurse specialist to deliver AT to one patient/patient–carer pair is estimated at £352 ($521).
Our findings show that a seven-week programme of AT significantly improved medication adherence and QoL compared with RC in a sample of medication non-adherent PD patients. Improvements were shown in the PDQ-39 overall score and in a range of sub-domains.
These findings suggest that AT may be useful for improving clinical outcomes in people with PD, as has been shown in other chronic conditions. AT in hypertension showed greater medication adherence significantly lowered blood pressure . When used in patients with schizophrenia, psychotic symptoms significantly improved . In both studies shifting beliefs about medication resulted in improved adherence behaviours. We only showed small improvements in beliefs, suggesting beliefs may not contribute significantly to non-adherence in PD. In contrast to other chronic conditions where beliefs about medication may play a key role, we found that most PD patients missed/mistimed doses because of forgetfulness and poor drug management, especially when cognitive impairment and/or regimen complexity presented. Parkinson's disease affects various cognitive processes (e.g. planning and problem-solving), potentially making drug management difficult . Furthermore, patients may be prescribed intricate regimens to manage symptoms and maximise ‘On’ time. This can further complicate treatment . Problem-solving is believed to help people develop meaningful strategies . Therefore, using AT to facilitate problem-solving may explain some of the observed improvements in adherence in PD.
The AT assessment (a structured interview) showed medication ambivalence, depressive episodes, denial concerning the indication for treatment and poor understanding of basic antiparkinsonian pharmacodynamics were all reasons for non-adherence in this study. As these factors were consistent with our previous findings , we incorporated our greater understanding of these issues into the AT intervention. Patients often mistimed doses as they were unaware of the reasons for correct time interval dosing. Facilitating an appreciation of pharmacodynamics, by visual reference to the effect of erratic dosing (using a simple graph to demonstrate dopamine level fluctuations), greatly enhanced adherence behaviours as patients could visualise how poor pill taking may affect their symptom control.
Adherence therapy had a significant effect on QoL. Because of the known variability in PD symptoms and the short follow-up duration, this was unexpected. We showed a significant improvement in mobility on the PDQ-39. As rigidity and bradykinesia are sensitive to antiparkinsonian medication , and therefore may be most likely to respond to treatment, it is reasonable to suggest that the improved adherence may account for the observed improvements in mobility. Despite showing no significant difference between the groups in MDS-UPDRS part IV, fewer intervention group participants reported motor complications than in the control group. Therefore, it is possible that patients who improve in adherence may also report experiencing fewer motor complications. We plan to assess this in future investigations.
Surprisingly, cognition improved in the PDQ-39. Although cognitive dysfunction is refractory to dopamine replacement, bradyphrenia is however levodopa responsive; previous research has shown improvement in neuropsychological tests following optimisation of dopamine replacement therapy [26, 27]. As our participants were not substantially cognitively impaired, it is possible that the identified improvements in cognition may result from participants feeling mentally more alert because of improved medication adherence. It is important to note however that this may be a chance finding which should be interpreted cautiously. A larger trial with a cognitive assessment as a predefined study outcome will provide evidence of the reliability of this finding.
The benefits observed in this trial have important clinical implications. Optimising adherence could be as essential as optimising dosage. Therefore, ensuring sound adherence prior to dose escalation is critical. This highlights the requirement to consider adherence when medications are reviewed in PD. Furthermore, acknowledging factors associated with non-adherence  may help identify non-adherent patients. AT is an intervention that can be delivered following a short training period. Ideally professionals should facilitate optimal adherence by utilising the principles of AT from diagnosis, especially as early PD is associated with non-adherence . Grosset et al. improved adherence in PD through a didactic educational approach . Whilst education may be effective for some, the diversity of factors associated with non-adherence suggests that people with PD may require a patient focused intervention such as that described in this study.
Our study has several strengths. Firstly, other than the specific requirement for medication non-adherence, our inclusion criteria were deliberately open. Drug trials often exclude PD patients because of older age (> 65), greater disease severity and impaired cognition . We purposely did not exclude such patients [mean age 71.9 (SD 8.9) years]. Therefore, our findings are generalisable to older people in which PD is prevalent. Unlike many adherence interventions, AT was delivered in participants' homes. Consequentially, severely affected patients were able to participate that could not have done so if required to attend hospital/clinic. Furthermore, deficits in set-shifting, defined as the process of updating cognitive strategies for changing environments/tasks , are prevalent and are associated with poor problem-solving in PD . Therefore, delivering AT at home, where set-shifting can be minimised by focusing on contextual tasks, may help patients develop strategies to facilitate optimal medication adherence. Home delivery also resulted in highly committed patients, emphasising the acceptability of the therapy.
This study was part of a PhD project where time and resources were limited. Limitations must therefore be acknowledged. Firstly, a longer follow-up period (e.g. 12 months) is desirable to evaluate whether improved adherence is sustained and how this correlates with clinical outcomes. In a larger trial of AT in PD, we will therefore assess adherence at 6 and 12-month follow up in order to provide adherence data which will inform a multi-centre investigation. Secondly, whether improvements in the intervention group resulted from the efficacy of AT or simply were because of increased patient–professional interaction (i.e. increased attention) requires further investigation. In our planned larger scale study with a longer follow up, an intervention (e.g. frequent sessions not related to medication adherence) that offers patient/professional interaction equal to that provided in the AT group will be delivered to the control group participants for comparison against AT.
Bias may have been introduced as AT was delivered by one therapist. Although outcomes were self-reported, the lack of blinding may have resulted in the intervention group participants over-reporting adherence to please the investigator. Also, because of potential bias we did not assess MDS-UPDRS part III (because of limited resources this would have been assessed by DJD who also delivered AT). Therefore, collection of all follow-up measures including an adherence assessment and MDS-UPDRS III by an individual masked to group allocation is planned in our larger scale trial of AT in PD. Despite showing a significant change in adherence from baseline to follow up, it is important to note that the 4-item MMAS scale is a relatively crude tool. In future trials, we intend to assess adherence using a variety of techniques such as Medication Electronic Monitoring Devices (MEMS caps), the current reported gold-standard method. Finally, our subgroup interaction analyses were underpowered. As the lack of a spouse/carer and symptoms of anxiety/depression were shown to be associated with non-adherence in PD , greater power may reveal an interaction between these factors and the treatment effect of AT.
In summary, we conclude that AT may be an effective, novel therapy for PD. Health professionals could be easily trained to utilise the core principles of AT as part of routine practice. The small NNT observed and the low cost of the intervention suggests AT may be cost effective, especially when considering the potential future cost savings of increased adherence. A larger pragmatic study is therefore required to examine the clinical and cost effectiveness of this intervention when delivered by multiple trained PD nurse specialists and when compared with a control group intervention. Whether the effectiveness of this short intervention is sustained over time also requires further investigation which our planned multi-centre RCT of AT in PD seeks to address.
DJD, PKM, KHOD, RJG, ABC, PW, MP and KS all contributed to the design of the study. DJD, PKM, KHOD and RJG drafted the paper. ABC conducted statistical analyses. All authors provided revisions of the paper and have approved the final manuscript. DJD is the guarantor.
Funding and acknowledgements
We wish to thank all participants for their contributions to this study. We thank Fiona Reading, Michelle Green, Rachael Rendell, Terri Johns, Bronwen Roper and the Trial Steering Committee lay members, Garth Ravenhill and Philip Harrison, for their invaluable support. We give special thanks to Debbie Davey for her valuable efforts as data clerk. This research was funded through a University of East Anglia PhD studentship.