Real-world adherence to, and persistence with, once- and twice-daily oral disease-modifying drugs in patients with multiple sclerosis: a systematic review and meta-analysis

Background Nonadherence to disease-modifying drugs (DMDs) for multiple sclerosis (MS) is associated with poorer clinical outcomes, including higher rates of relapse and disease progression, and higher medical resource use. A systematic review and quantification of adherence and persistence with oral DMDs would help clarify the extent of nonadherence and nonpersistence in patients with MS to help prescribers make informed treatment plans and optimize patient care. The objectives were to: 1) conduct a systematic literature review to assess the availability and variability of oral DMD adherence and/or persistence rates across ‘real-world’ data sources; and 2) conduct meta-analyses of the rates of adherence and persistence for once- and twice-daily oral DMDs in patients with MS using real-world data. Methods A systematic review of studies published between January 2010 and April 2018 in the PubMed database was performed. Only studies assessing once- and twice-daily oral DMDs were available for inclusion in the analysis. Study quality was evaluated using a modified version of the Newcastle-Ottawa Scale, a tool for assessing quality of observational studies. The random effects model evaluated pooled summary estimates of nonadherence. Results From 510 abstracts, 31 studies comprising 16,398 patients with MS treated with daily oral DMDs were included. Overall 1-year mean medication possession ratio (MPR; n = 4 studies) was 83.3% (95% confidence interval [CI] 74.5–92.1%) and proportion of days covered (PDC; n = 4 studies) was 76.5% (95% CI 72.0–81.1%). Pooled 1-year MPR ≥80% adherence (n = 6) was 78.5% (95% CI 63.5–88.5%) and PDC ≥80% (n = 5 studies) was 71.8% (95% CI 59.1–81.9%). Pooled 1-year discontinuation (n = 20) was 25.4% (95% CI 21.6–29.7%). Conclusions Approximately one in five patients with MS do not adhere to, and one in four discontinue, daily oral DMDs before 1 year. Opportunities to improve adherence and ultimately patient outcomes, such as patient education, medication support/reminders, simplified dosing regimens, and reducing administration or monitoring requirements, remain. Implementation of efforts to improve adherence are essential to improving care of patients with MS.


Systematic literature review
A systematic literature search was performed of all studies published between January 2010 and April 2018 in the PubMed database that evaluated adherence or persistence to oral DMDs. The search strategy used the following terms: (Aubagio OR cladribine OR dimethyl fumarate OR fingolimod OR Gilenya OR Tecfidera OR teriflunomide OR oral OR disease modifying drug OR DMD OR disease modifying therapy OR DMT) AND multiple sclerosis AND (adherence OR compliance OR persistence OR discontinuation). A priori exclusion criteria were: lack of primary data; lack of primary realworld DMD adherence/persistence data; lack of oral DMD adherence/persistence data; pediatric studies; non-English studies; and abstract-only available. Two reviewers independently reviewed the search results and reference lists of selected articles to identify additional appropriate studies and carried out data extraction. Full search strategy and search results are provided in Additional File 1.
Information extracted from the screened articles included the type of study/data source; study population (baseline demographic and clinical characteristics); treatment arms; duration of follow-up after DMD initiation; sample size; outcomes evaluated (including definition of adherence and method of measurement); clinical results; secondary results; and strengths and limitations of the studies.
The quality of selected studies was evaluated using a modified version of the Newcastle-Ottawa Scale (NOS), a tool for assessing the quality of observational studies [20]. The NOS was modified as it was primarily designed to evaluate case-control and comparative studies. Study quality was evaluated with the modified NOS from the perspective of study design and patient selection as well as outcome to generate an overall assessment of the quality of the studies and their internal validity. Three criteria were evaluated under the NOS 'study design and patient selection' perspective: ascertainment of the intervention/validity of study design, patient selection, and outcome not present at the start of the study. Criteria evaluated under the 'outcome evaluation' perspective were appropriateness of the measure of adherence/ persistence, adequacy or appropriateness of duration of follow-up, and thoroughness of follow-up for all patients. Each criterion was given a full-, partial-, or poorquality score (Table 1). Per Cochrane Collaboration [21] and other recommendations [22] that meta-analyses not be adjusted on the basis of quality, the results of the study quality assessment were presented as standard tables and systematic narrative description and commentary about each of the elements.

Meta-analyses
The selection of endpoints was driven by the availability of data in the published, peer-reviewed literature. Adherence was evaluated using either the medication possession ratio (MPR) or the proportion of days covered (PDC). MPR was calculated as the total number of days of medication supply between the first prescription claim and the last prescription claim issued during the followup period divided by the total number of days in the follow-up period. A variable follow-up period was used for the MPR denominator (number of days between the index date and the last prescription dispensed inclusive of supply) [23]. PDC was calculated as the total number of days in the follow-up period in which medication was available (excluded overlapping days' supply) divided by the duration of the observation period (i.e., 1 year in the case of oral DMD adherence studies). Adherence was calculated as means (mean MPR and mean PDC) and the proportion of patients attaining the 0.8 threshold, which is commonly considered an acceptable level of adherence [24]. Discontinuation was defined as the proportion of patients who either switched DMDs or discontinued DMD medications altogether. Analyses were conducted for 5 separate endpoints over a 1-year follow-up period: (1) mean MPR for patients overall (2); mean PDC for patients overall (3); proportion of patients 'adherent', defined as proportion with MPR ≥80% (4); proportion of patients 'adherent', defined as proportion with PDC ≥80%; and (5) proportion of patients who discontinued the initial treatment.
In line with published recommendations regarding the use of real-world data in meta-analyses, [25] statistical heterogeneity was evaluated using Cochran's Q test and the I 2 statistic, which provide a measure of the presence of heterogeneity and the share of dispersion across studies, respectively [26]. If Q was significant and I 2 was > 50%, the random-effects model (REM) was used to calculate pooled summary estimates; otherwise, a fixedeffect model was used. The studies were weighted according to the extent of variation among the intervention effects. Egger's test was used to detect publication bias [27]. The 'metaprop' and the 'metamean' functions in the R statistical programming language [28] were used to conduct the meta-analysis, and a p-value of < 0.05 was considered statistically significant. Subgroup analyses conducted included study location (i.e., US/ex-US studies) and study design (i.e., prospective cohort vs. retrospective chart review vs. administrative claims database evaluation). The impact of any single study on the overall results was assessed using a leave-one-out sensitivity analysis, in which each study was iteratively removed and the findings compared to the overall meta-analysis.

Systematic literature review
From a total of 510 abstracts identified, 31 studies comprising 16,398 patients were included in the systematic review after applying exclusion criteria (Fig. 1). Table 2 provides information about the individual studies. No studies evaluating cladribine tablets were identified due to its recent approval, therefore the analyses focused on once-and twice-daily oral maintenance DMDs. Siponimod was not available at the time of the study and is not included in the analyses.
Quality assessments of the selected studies are presented in Table 1. For the ascertainment of the intervention/validity of study design criteria in the study design and patient selection perspective, approximately half of the studies had a full-quality score and half had a partial-quality score. One study (Lapierre et al. 2016) was assigned a poor-quality score for patient selection because the patient population was neither wellcharacterized nor representative of patients with RRMS [29]. For the criterion of outcome of interest not present at the start of the study, poor-quality scores were assigned to administrative claims database analyses ) as they were unable to ascertain patient prescription abandonment [5,11,12,17,[30][31][32]. For the three criteria in the Abbreviations: RRMS relapsing-remitting multiple sclerosis, • full-quality score, • partial-quality score, ○ poor-quality score For the ascertainment of the intervention/validity of study design criterion, if the study was a medical chart review, evidence that there was an effort made to validate reported data resulted in a full-quality score. Otherwise, the medical chart review or registry study was assigned a partial-quality score. Prospective cohort studies and administrative claims database evaluations were given a full-quality score for this criterion For the patient selection criterion, studies were given a full-quality score if the patients were well characterized (i.e., age, sex, region, duration of MS diagnosis, MS severity, prior treatments, current treatment) and were representative of patients with RRMS. Studies were not penalized for including selected populations or for only evaluating a single center because it was felt that these studies were still valid cohort studies. Studies were given a poor-quality score if the patient population was not well-characterized and was not representative of patients with RRMS For the outcome of interest not present at the start of the study criterion, studies were given a full-quality score if they attempted to capture patient prescription abandonment and thoroughly described how it was ascertained. Administrative claims database analyses were not able to ascertain this, and were therefore given a poor-quality score for this criterion For the appropriate measurement of adherence/persistence criterion, studies with a full-quality score had to appropriately define and measure adherence and persistence and include and/or delineate switching for discontinuation The adequate/appropriate duration of follow-up criterion required full-quality studies to measure adherence/persistence over 1 year as this was the most common time horizon used and enabled comparability The all patients accounted at follow-up criterion required that all patients evaluated were followed up throughout the study and did not have missing data
Subgroup analyses were only conducted for the outcome of discontinuation due to the small number of studies reporting MPR and PDC oral DMD adherence. When studies were analyzed by US and ex-US groupings, similar proportions of patients were found to discontinue DMD therapy ( (Fig. 5a). The proportion of patients discontinuing DMD therapy was greatest for studies evaluating administrative claims databases (29.0, 95% CI 22.0-37.2%), followed by prospective cohort studies (24.7, 95% CI 18.7-31.9%) and medical chart reviews (22.9, 95% CI 18.4-28.1%) (Fig. 5b), though some overlapping of CIs was apparent. The results of the leave-one-out sensitivity analysis confirmed that removal of individual studies did not affect the results (Supplementary Figs. 1, 2a, and 2b).

Discussion
This is the first meta-analysis to assess real-world adherence to and persistence with multiple oral maintenance DMDs in patients with MS. Results showed that one in five patients do not adhere to once-or twice-daily oral maintenance DMDs, and one in four patients discontinue the initially-prescribed oral DMDs before 1 year. Ninety-five percent CIs for the estimates of adherence and persistence were wide, reflecting the heterogeneity in rates of adherence and discontinuation.
Adherence to DMDs is an important aspect of optimizing patient care in MS as greater adherence has been shown to be associated with improvements in relapse outcomes and quality of life, fewer hospitalizations and emergency room visits, decrease in neuropsychological issues, fewer days of work lost, and lower MS-related medical costs [11][12][13][14][54][55][56][57]. Compared with nonadherence, adherence to newly initiated DMDs (oral or injectable) over 1 year among patients in the US was associated with a decrease of 42% in the likelihood of relapse, 38% in    emergency visits, and 52% in hospitalizations, as well as an average of 0.7 fewer outpatient visits annually (all p < 0.0001) [11]. Based on the differences in predicted US mean costs over the 12-month post-initiation period, adherence was estimated to decrease the total annual medical care costs by $5816 per patient (all costs are reported in 2015 US$), including outpatient visits ($2802), emergency visits ($171), and hospitalizations ($1953) compared with nonadherence [11]. Persistence with DMDs for MS is also important for achieving the best clinical outcomes, as DMD persistence has been shown to be associated with decreased likelihoods of inpatient admission or emergency room visits, [58] decreased relapses and disease progression, [59] and lower costs [60]. An increased understanding of barriers to DMD adherence and persistence and the implementation of efforts to improve adherence and persistence are important. Realworld data on adherence to DMDs in patients with MS can help inform therapeutic decision making [61]. Factors that generally influence adherence and persistence in chronic illness, particularly when cognition is a factor (as in MS), include the frequency and complexity of the dosing regimen, [62,63] the need for ongoing safety monitoring, office visits, or waiting time associated with dosing, [64] prior treatment experience, [64] and the presence or absence of active symptoms at the time of dosing [64]. Studies in other therapeutic areas have shown that simpler and less frequent dosing produces greater adherence than more frequent administration [62,63,65]. A study of MS patient preferences for oral treatments also found that the most important driver of predicted nonadherence was frequency of daily dosing (17.4% out of 100%) [66].
Depression has also been shown to be associated with decreased DMD adherence and persistence in patients with MS. Gerber, et al. and Munsell, et al. both reported a significant relationship between comorbid depression and nonadherence, [12,17] whereas Lattanzi, et al. reported a significant relationship between depression and persistence [35]. Previous studies have shown that patients with MS and comorbid depression are significantly more likely to be nonadherent to DMDs than patients with MS without comorbid depression [67,68].
The rates of adherence and discontinuation may differ among oral DMDs due to several potential Fig. 2 Meta-analysis of mean adherence rate as determined by a) MPR or b) PDC. Note: For studies for which results for treatment-naive and treatment-experienced patients were reported separately (combined data were not available), data were combined; for studies reporting data for more than 1 oral DMD (combined data were not reported), data were combined; for studies reporting data for subgroups (combined data were not reported), data were combined. The area of each grey square is proportional to the study's weight in the meta-analysis. Weight values are rounded. Abbreviations: CI, confidence interval; DMD, disease-modifying drug; MPR, medication possession ratio; MRAW, raw mean; PDC, proportion of days covered factors such as the dosing regimen (e.g. once-versus twice-daily), efficacy, tolerability, and adverse events [15,35,36]. Assessments of the specific impacts of these factors could not be conducted in the current study due to the small number of studies evaluating the individual oral DMDs' adherence/persistence over 1 year (i.e., 7 studies evaluated dimethyl fumarate, 10 studies evaluated fingolimod, and only 2 studies evaluated teriflunomide) and due to the challenges in elucidating the specific reasons for any associations that might be observed among studies with significant heterogeneity.
This study was conducted in line with recommendations available in the literature for the use of real-world evidence in meta-analyses [25]. In our findings, 95% CIs showed a wide range of values, particularly for the proportion of patients who were adherent, but this is expected when pooling observational (real-world) data [25]. Heterogeneity is likely to arise because of differences in patient populations, treatments, study design, outcomes, and data quality [69]. This was evident in the results of our assessment of the quality of the studies, which highlighted how the appraisal of the studies needed to be adapted for their individual design (hence our modification of the NewCastle Ottawa Scale [NOS]). Study location, in and of itself, was not a source of heterogeneity in the current meta-analysis.
The I 2 statistics obtained in this study ranged from 93.8 to 99.5%. These values are consistent with those noted in other published meta-analyses of medication adherence across indications [19,[70][71][72][73].
Subgroup analyses demonstrated that there were essentially no differences between US-and ex-US-based studies and support the need for a better understanding of why patients worldwide discontinue oral DMD treatment. The study design subgroup analyses showed overlapping CIs, indicating a lack of significant differences. However, numerical differences suggested that administrative claims data analyses may more fully capture discontinuation than prospective cohort studies and medical chart reviews. With prospective cohort studies and medical chart reviews, bias may arise from patient or investigator report, whereas administrative claims database analyses provide objective billing information for medication dispensed. Also, cohort studies may also inherently encourage patients to remain in the study and continue treatment. Chart reviews may not capture all discontinuations, depending upon the availability and quality of follow-up data. Further subgroup analyses were limited by the small number of real-world/ Fig. 3 Meta-analysis of proportion of patients adherent to a DMD as determined by MPR or PDC. Note: For studies for which results for treatment-naive and treatment-experienced patients were reported separately (combined data were not available), data were combined; for studies reporting data for more than 1 oral DMD (combined data were not reported), data were combined; for studies reporting data for subgroups (combined data were not reported), data were combined. The area of each grey square is proportional to the study's weight in the meta-analysis. Weight values are rounded. Abbreviations: CI: confidence interval; DMD: disease-modifying drug; MPR: medication possession ratio; PDC: proportion of days covered observational studies, suggesting that more real-world research is needed. This is the second published meta-analysis of realworld adherence to or persistence with oral DMDs in patients with MS. Kantor et al. 2018 conducted a systematic review and meta-analysis focusing on the realworld persistence with fingolimod in patients with RRMS [19]. In contrast, our meta-analysis evaluated both adherence and persistence in all once-and twicedaily oral maintenance DMDs. Unlike the Kantor  There are limitations to this study. Although nonrandomized cohort studies or observational studies may provide a more 'real-world' representation of outcomes, costs, and utilization, differences in baseline characteristics can introduce biases, and the influence of unmeasured factors cannot be ruled out. As described, the substantial proportion of heterogeneity found across the studies is also a limitation. The use of a variable follow-up period for the MPR denominator potentially contributed to an inflated mean MPR [23]. Additionally, MPR may overestimate medication adherence as compared to PDC, as it counts the total number of days of medication supply, which may be inflated by patients who fill their prescriptions early and gain extra supply within a period [74]. No studies evaluating cladribine tablets were identified due to its recent approval, therefore the analyses focused on once-and twicedaily oral maintenance DMDs. Siponimod was not available at the time of the study and is not included in the analyses. Finally, the limited number of studies restricted the ability to analyze MPR and PDC adherence in greater detail and the ability to perform subgroup analysis across adherence and persistence measures.

Conclusions
Adherence to treatment is an important issue for the management of patients with MS. This meta-analysis of real-world studies showed that approximately one in five Fig. 4 Meta-analysis of proportion of patients discontinuing a DMD. Note: For studies for which results for treatment-naive and treatmentexperienced patients were reported separately (combined data were not available), data were combined; for studies reporting data for more than 1 oral DMD (combined data were not reported), data were combined; for studies reporting data for subgroups (combined data were not reported), data were combined; for Lattanzi et al., number of patients discontinuing at 12 months included patients for whom data were not available at 12 months because they stopped taking medications  Subgroup meta-analyses of proportion of patients discontinuing a DMD. Note: For studies for which results for treatment naive and treatment-experienced patients were reported separately (combined data were not available), data were combined; for studies reporting data for more than 1 oral DMD (combined data were not reported), data were combined; for studies reporting data for subgroups (combined data were not reported), data were combined. The area of each grey square is proportional to the study's weight in the meta-analysis. Weight values are rounded. Abbreviations: ACD, administrative claims database; CI, confidence interval; MCR, medical chart review; PC, prospective cohort patients with MS do not adhere to once-or twice-daily oral maintenance DMD treatment regimens, and one in four patients with MS discontinue once-or twice-daily oral maintenance DMDs before 1 year. Wide 95% CIs for the estimates of adherence and discontinuation reflect the heterogeneity in the rates that was observed. Opportunities to address barriers to DMD adherence in patients with MS remain, such as patient education efforts to manage expectations and to emphasize the importance of adherence, implementation of medication support/reminder techniques, simplification of dosing regimens, and reducing the need for specialized administration or monitoring requirements [61]. Implementation of efforts to improve adherence are essential to improving care of patients with MS.

Supplementary information
Supplementary information accompanies this paper at https://doi.org/10. 1186/s12883-020-01830-0. Additional file 2: Supplementary Figure 1. Leave-one-out sensitivity analysis for the proportion of patients discontinuing a DMD. For studies for which results for treatment naive and treatment-experienced patients were reported separately (combined data were not available), data were combined; for studies reporting data for more than 1 oral DMD (combined data were not reported), data were combined; for studies reporting data for subgroups (combined data were not reported), data were combined. Abbreviations: DMD, disease-modifying drug.
Additional file 3: Supplementary Figure 2. Leave-one-out sensitivity analysis for subgroup analyses for the proportion of patients discontinuing a DMD. For studies for which results for treatment naive and treatment-experienced patients were reported separately (combined data were not available), data were combined; for studies reporting data for more than 1 oral DMD (combined data were not reported), data were combined; for studies reporting data for subgroups (combined data were not reported), data were combined. Abbreviations: DMD: diseasemodifying drug.