Biweekly online surveys were collected over 1 year in 112 people with IBD. The MARS-5 scores were obtained at weeks 0 and 52. The MARS-5 was significantly associated with adherence based on medication monitoring data at baseline (r = 0.48) and week 52 (r = 0.57). Sensitivity and specificity for adherence ≥80% and ≥90% were maximized at MARS-5 scores of >22 and >23, respectively.
Abstract
Introduction
We aimed to validate the Medication Adherence Report Scale-5 (MARS-5) as a tool for assessing medication adherence in inflammatory bowel disease (IBD) and to determine predictors of medication adherence.
Methods
One hundred twelve (N = 112) adults with confirmed IBD participating in the longitudinal Manitoba Living With IBD Study were eligible. Demographics, IBD type, surgeries, disease activity (using the Inflammatory Bowel Disease Symptom Inventory and fecal calprotectin levels), perceived stress, and medication use were collected biweekly through online surveys. The MARS-5 scores were obtained at baseline and at 1 year. Correlation between medication monitoring data and MARS-5 scores was performed and the optimal MARS-5 cutoff point for adherence assessment determined. Predictors of medication adherence were assessed at both ≥90% and ≥80%.
Results
Participants were predominantly female (71.4%), mean age was 42.9 (SD = 12.8), and the majority (67.9%) had Crohn disease (CD). Almost half (46.4%) were taking more than 1 IBD medication, with thiopurines (41.9%) and biologics (36.6%) the most common. Only 17.9% (n = 20) were nonadherent at a <90% level; of those, 90% (n = 18) were using oral medications. The MARS-5 was significantly associated with adherence based on medication monitoring data at baseline (r = 0.48) and week 52 (r = 0.57). Sensitivity and specificity for adherence ≥80% and ≥90% were maximized at MARS-5 scores of >22 and >23, respectively. Having CD (OR = 4.62; 95% confidence interval, 1.36-15.7) was the only significant predictor of adherence.
Conclusion
MARS-5 is a useful measure to evaluate adherence in an IBD population. In this highly adherent sample, disease type (CD) was the only predictor of medication adherence.
INTRODUCTION
Inflammatory bowel diseases (IBD) are chronic medical conditions, which untreated can result in a high degree of morbidity, disability, and cost to the health care system.1-6 There have been significant advances in pharmacological treatments for IBD, considering both the agents (eg, biologics) and the approach to medication use, with outcomes demonstrating high relapse rates in patients who discontinue therapy.6–8 Thus, it is important that individuals with IBD are highly adherent to their prescribed medications to reduce the risk of flares and the long-term complications associated with the disease.
It has previously been reported that medication adherence, defined as the extent to which patients take medications as prescribed by their health care provider, is greater in acute illness and significantly decreases in chronic disease.9 Factors associated with both medication adherence and nonadherence have been extensively studied across a wide range of diseases.5, 9-15 In the IBD population, these include a variety of biopsychosocial factors, such as age, sex, disease type, psychological illness, physician-patient communication, and medication beliefs.5, 11, 14, 16 However, there are inconsistencies across studies when reporting factors associated with adherence and rates of adherence.2, 5, 10, 17 A large meta-analysis examining medication adherence in IBD reported variable adherence rates, ranging from 7 to 72%, with a variety of adherence measurement approaches being utilized.10 Furthermore, obtaining an accurate medication adherence history can be difficult for the prescribing physician because there is not a single validated method that is universally used for all diseases and some patients may have difficulty acknowledging lapses directly.10, 18
There is variation in the use of adherence measurement within the IBD population, with studies variously using the Morisky Medication Adherence Scale, the Visual Analogue Scale, the Beliefs About Medicine Questionnaire, and more recently the Medication Adherence Report Scale-4 and the Medication Adherence Report Scale-5 (MARS-5).1-4, 10, 11, 19 Most of these measures, although used extensively, have not been validated in patients with IBD. Other measures to determine adherence have been used, such as drug metabolite levels and prescription refill rates; however, these are limited by availability, cost, and reporting bias.20 Because medication adherence in the clinic setting is often determined through physician-patient interactions and questioning, as compared to a validated questionnaire or scale, it can be difficult to determine the true medication adherence and to accurately predict which variables play a role in adherence.9, 10
The MARS-5 has been used as a measurement of adherence in many chronic diseases, including chronic obstructive pulmonary disease, stroke, and type 2 diabetes.12, 15, 21 Whereas the MARS-5 has been used in IBD research studies previously, it has never been validated as an accurate measurement tool in this population.11 Research has shown that people find estimating adherence to be simpler than giving exact numbers, and given the simplicity, low cost, and speed of self-report questionnaires like the MARS-5, these adherence tools have been recommended for use in a clinical setting.15, 22
The aim of the current study was to examine medication adherence in a cohort of adults with IBD in Manitoba, Canada, comparing biweekly medication monitoring data to the MARS-5 in order to understand the utility of the scale for IBD. We also explored the effects of IBD symptoms, psychological stress, and disease activity in our population to determine predictors of adherence.
METHODS
Study Population and Characteristics
One hundred fifty-five patients participating in the Manitoba Living With IBD Study were eligible for the current study. The Manitoba Living With IBD Study, reported in detail elsewhere,23 enrolled adults with previously active IBD symptoms within the prior 2 years and followed them prospectively with online survey questionnaires every 2 weeks for 1 year. Data were obtained on disease course, psychological functioning, health comorbidities, and medication use. Participants for the Manitoba Living with IBD Study were recruited from a previous longitudinal cohort study, regional gastroenterology clinics, a provincial research registry, and the IBD Clinical and Research Center website (ibdmanitoba.org), and through study advertisements via posters in Manitoba hospitals and gastroenterologist offices.23 After excluding Manitoba study participants not taking any IBD-related medications (n = 16), those taking medications on an as-needed basis (n = 8), those who completed <50% of the biweekly medication reporting data (n = 16), and those who did not complete the study (n = 3), 112 patients (72% of the sample) were included in the current study.
Descriptive data on demographics, IBD subtype, abdominal surgical history, IBD-related medications, fecal calprotectin levels (FCAL), IBD disease activity [Inflammatory Bowel Disease Symptom Inventory (IBDSI)], and perceived stress [measured by the Cohen Perceived Stress Scale (CPSS)] were collected at study baseline. The IBDSI is a validated questionnaire that reflects disease activity through a 36-question (long-form) and 24-question (short-form) survey questionnaire, which has previously been shown to have very good sensitivity and specificity when compared to the Harvey-Bradshaw Index [Crohn disease (CD)] and the Powell-Tuck Index [ulcerative colitis (UC)].24 The cutoffs for active symptomatic disease are >13 in UC and >14 in CD. Medication adherence was assessed by the MARS-5 (Table 1) at weeks 0 and 52 and through biweekly medication tracking that provided percentage adherence information over the preceding 2 weeks. The 10-item CPSS was collected at weeks 0, 26, and 52. A score of >8 indicated higher perceived stress. Disease activity was determined through intestinal inflammation, measured by stool samples for FCAL (active inflammation defined as >250 µg/g) collected at week 0, 26 and 52, and the IBDSI, reported through the online survey every 2 weeks, with the total 52-week scores summed and then averaged over 26 weeks for an average disease activity score over the duration of the 1-year study period.23
TABLE 1.
MARS-5.
| Item | Always | Often | Sometimes | Rarely | Never |
|---|---|---|---|---|---|
| “I forget to take them” | 1 | 2 | 3 | 4 | 5 |
| “I alter the dose” | 1 | 2 | 3 | 4 | 5 |
| “I stop taking them for a while” | 1 | 2 | 3 | 4 | 5 |
| “I decide to miss out a dose” | 1 | 2 | 3 | 4 | 5 |
| “I take less than instructed” | 1 | 2 | 3 | 4 | 5 |
The MARS-5 score was calculated by summing the numeric score (range 1-5) from each question for out of 25 (range 5-25). A higher score indicates better adherence
MARS-5
The MARS-5 assesses a patient’s typical medication adherence through 5 questions (eg, “I forget to take my medication”; “I alter the dose of my medication”), using a 5-level response format (1—always, 2—often, 3—sometimes, 4—rarely, and 5—never). Responses are summed for a total score ranging between 5 and 25, with higher scores indicating a higher level of adherence. Using the MARS-5, a score of ≥20 has previously been used as a cutoff for adherence in an IBD population, while other chronic diseases have defined adherence at a cutoff as high as 25.11-13, 15, 21
Biweekly Medication Monitoring
Participants provided information on prescribed medications biweekly online. Medication utilization was determined by asking participants to enter the percentage they had taken of each prescribed IBD related medication over the preceding two-week period. The two-week percentages for each IBD medication were averaged over one year to determine an overall adherence percentage. For participants prescribed multiple IBD medications, adherence data for each medication was averaged over the year, and then all prescribed IBD medications were averaged to give a total medication adherence value. For example, a participant with an adherence of 50% for azathioprine and 0% for adalimumab would have 25% adherence during the two-week period. If this participant did not change their adherence throughout the year, and continued to report 25% of their prescribed medications at each biweekly medication monitoring survey, their 1 year average would be 25%.
Adherence was defined in 2 ways: by adherence rates ≥90% and by adherence rates ≥80%, and nonadherence was defined in 2 ways: rates <90% and rates <80%. Any IBD-related medication that was prescribed for durations of longer than 4 weeks was included in the calculation of average adherence over the duration of the study. To better assess participants’ overall adherence to their medication, only those who had medication data for at least 50% (26 weeks) of the survey period were included in the primary study analysis. This cutoff was deemed minimally acceptable by the authors. A subgroup analysis for those with <50% biweekly medication adherence data was then performed to determine the effect of including these participants as either adherent (≥90%) or nonadherent (<90%).
Analysis
Unadjusted (one factor at a time) and adjusted (multiple factors in a model) logistic regression models were used to test the factors associated with medication adherence in an IBD population, using active IBD symptoms (IBDSI), active inflammation (FCAL), age, sex, and perceived stress (CPSS) as covariates. Two outcomes were assessed in these models: ≥80% adherence and ≥90% adherence. Odds ratios (ORs) and 95% confidence intervals (95% CI) are reported. Pearson correlation coefficients were then used to determine the correlation between MARS-5 and medication monitoring adherence data. Finally, we conducted receiver operating curves analysis to explore the optimal MARS-5 cutoff and the sensitivity and specificity of MARS-5 (using the respective cutoff) as a predictor of medication adherence. Given our highly adherent cohort, analysis was performed for adherence ≥80%, the generally accepted cutoff to define adherent behavior, and for adherence ≥90%, selected as a more rigorous threshold, for all 112 participants included in the study. To determine the effect of participants who had provided insufficient medication monitoring data (ie, <50%) and adherence data and were initially excluded in our reported results, we performed further multivariable logistic regression analysis that included this subgroup, defining them collectively in separate analyses as either adherent (≥90%) or nonadherent (<90%).
RESULTS
Participant Demographics, Medication Data, and Disease Activity
For the 112 participants, the mean age of the cohort was 42.9 (SD = 12.8) with a female predominance (71.4%). The majority of participants had a diagnosis of CD (67.9%) followed by UC (28.6%), and 4 (3.6%) were IBD type unclassified. A total of 37.5% (n = 42) participants had previously had at least 1 abdominal surgery before the start of the study, while more than one-fifth had 2 or more. Nearly 85% (n = 94) of the cohort were nonsmokers, with 11.7% (n = 13) smoking daily. Every participant was on at least 1 IBD-related medication, with almost half (46.4%) taking 2 IBD-related medications and nearly one-quarter taking 3 or more (n = 27; 24.1%) at some point over the course of the study (Table 2).
TABLE 2.
Characteristics of the Study Cohort at Baseline (N = 112)
| Characteristic | Result |
|---|---|
| Age, y, mean (SD) | 42.9 (12.8) |
| Sex, female n (%) | 80 (71.4) |
| Diagnosis | |
| CD, n (%) | 76 (67.9) |
| UC, n (%) | 32 (28.6) |
| Indeterminate, n (%) | 4 (3.6) |
| Abdominal surgery, n (%) | 42 (37.5) |
| ≥2 surgeries, n (%) | 23 (20.5) |
| Smoking | |
| Daily, n (%) | 13 (11.7) |
| Occasional, n (%) | 5 (3.6) |
| None, n (%) | 94 (84.7) |
| Number of IBD medications | |
| 1, n (%) | 33 (29.4) |
| 2, n (%) | 52 (46.4) |
| ≥3, n (%) | 27 (24.1) |
Table 3 reports the medication data, IBDSI symptom scores, and MARS-5 adherence scores for the cohort. Thiopurines were the most common medication, prescribed and taken by 41.9% (n = 47) of participants. There were 36.6% (n = 41) on an infusion biologic, with another 21.4% (n = 24) on adalimumab subcutaneously, and 5-aminosalycylate formulations were taken by 27.6% (n = 31). The mean IBDSI score across the cohort over the study period was 15.5 (SD = 9.2), indicating active IBD symptoms (Table 3). Only 10.7% (n = 12) of participants had a MARS-5 value <20, a previously used cutoff for nonadherence in IBD,11 with approximately half the cohort (45.5%) reporting a MARS-5 score <23 (Table 3).
TABLE 3.
Medication Data, Disease Activity, MARS-5, and CPSS
| Characteristic | Result |
|---|---|
| IBD medication class* | |
| Oral 5-ASA, n (%) | 31 (27.6) |
| Rectal 5-ASA, n (%) | 10 (8.9) |
| Adalimumab s.c., n (%) | 24 (21.4) |
| Infusion biologic, n (%) | 41 (36.6) |
| Thiopurine, n (%) | 47 (41.9) |
| Methotrexate, n (%) | 4 (3.5) |
| Budesonide, n (%) | 4 (3.5) |
| Oral steroid, n (%) | 12 (10.7) |
| Antibiotic, n (%) | 4 (3.5) |
| IBDSI score, mean (SD)† | 15.5 (9.2) |
| MARS-5 score, mean (SD) | 22.5 (2.2) |
| MARS-5 week 0, mean (SD) | 22.3 (2.5) |
| MARS-5 week 52, mean (SD) | 22.7 (2.5) |
| MARS-5 <20, n (%) | 12 (10.7) |
| Monitoring adherence < 90%, n (%)‡ | 20 (17.9) |
| CPSS score, mean (SD)† | 5.9 (2.7) |
5-ASA indicates 5-aminosalicylate.
*Number of patients taking class of medication at any time throughout study period (1 year).
†Mean over study period (1 year).
‡Number of patients reporting nonadherence (<90%) over study period (1 year).
s.c., subcutaneous.
Adherence ≥90%
Adherence rates based on medication monitoring varied for each medication class (Table 4). Ninety-two participants (82.1%) were defined as adherent. Twenty participants (17.9%) were nonadherent; 90% (n = 18) of those were taking only oral medications, accounting for 17.6% of all oral medication users (Table 4). Only 1 infusion biologic user (2.4% of all biologic users) and 1 adalimumab user (4.2% of all adalimumab users) were classified as nonadherent.
TABLE 4.
Nonadherence Based on Medication Monitoring, by Medication Class
| Medication Type | Adherence <90% |
|---|---|
| Overall nonadherence in entire cohort (n = 112) | 17.9% (n = 20) |
| Oral medication (n = 102) | 17.6% (n = 18) |
| Infusion biologic (n = 41) | 2.4% (n = 1) |
| Adalimumab s.c. (n = 24) | 4.2% (n = 1) |
| Adherence <80% | |
| Overall nonadherence in entire cohort (n = 112) | 9.8% (n = 11) |
| Oral medication (n = 102) | 10.7% (n = 11) |
| Infusion biologic (n = 41) | 0% (n = 0) |
| Adalimumab s.c. (n = 24) | 0% (n = 0) |
Nonadherence calculated as percentage of medications taken relative to medications prescribed.
s.c., subcutaneous.
The results of the logistic regression analyses are reported in Table 5, adjusted for active IBD symptoms (IBDSI), active inflammation (FCAL), age, sex, disease type, and perceived stress (CPSS). A diagnosis of CD (OR = 4.62; 95% CI, 1.36-15.7; P = 0.01) was the only significant predictor of medication adherence. Baseline active symptomatic disease (IBDSI; OR = 1.04; 95% CI, 0.35-3.11; P = 0.72), baseline active inflammation (FCAL; OR = 0.44; 95% CI, 0.13-1.45; P = 0.14), baseline high perceived stress (CPSS; OR = 0.67; 95% CI, 0.19-2.32; P = 0.50), age >55 (OR = 2.37; 95% CI, 0.65-8.65; P = 0.48), and female sex (OR = 0.39; 95% CI, 0.10-1.53; P = 0.15) were not shown to have any predictive value of adherence ≥90%.
TABLE 5.
ORs and 95% CI for Predictors of Adherence Based on Medication Monitoring Adjusted for Active Inflammation (FCAL), Active IBD Symptoms (IBDSI), Disease Type, Sex, Perceived Stress, and Age
| Variable | Adjusted Regression Adherence >90% OR (95% CI)† | P* | Unadjusted Regression Adherence >90% OR (95% CI) | P* | Unadjusted Regression Adherence >80% OR (95% CI) | P* |
|---|---|---|---|---|---|---|
| Active inflammation (FCAL >250 µg/g) | 1.04 (0.35-3.11) | 0.72 | 0.84 (0.32-2.2) | 0.72 | 1.58 (0.44-5.75) | 0.48 |
| High perceived stress (CPSS >8) | 0.67 (0.19-2.32) | 0.50 | 0.69 (0.24-2.01) | 0.50 | 1.48 (0.30-7.31) | 0.63 |
| Active IBD symptoms IBDSI >13, UC; >14, CD | 0.44 (0.13-1.45) | 0.14 | 0.45 (0.16-1.30) | 0.14 | 0.63 (0.17-2.40) | 0.50 |
| CD | 4.62 (1.36-15.7) | 0.01 | 2.58 (0.93-7.15) | 0.07 | 4.15 (1.09-15.9) | 0.04 |
| Female | 0.39 (0.10-1.53) | 0.15 | 0.38 (0.10-1.41) | 0.15 | 0.23 (0.03-1.84) | 0.17 |
| Age, y (ref <45): | ||||||
| 45-54 | 3.08 (0.62-15.29) | 0.17 | 2.10 (0.54-8.22) | 1.63 (0.25-10.58) | ||
| ≥55 | 2.37 (0.65-8.65) | 0.19 | 1.86 (0.61-5.70) | 0.27 | 1.06 (0.24-4.61) | 0.94 |
| Smoking | 1.22 (0.22-6.87) | 0.23 | 1.77 (0.37-8.47) | 0.47 | 1.90 (0.22-15.93) | 0.55 |
*P values in bold represent statistical significance.
†C-statistic = 0.712.
When we ran the same models but included the 16 people with missing adherence data, assuming that they either all had ≥90% adherence or all had <90% adherence, our results were almost identical to those described above. Specifically, CD was the sole predictor of adherence (data not shown).
Adherence ≥80%
Further subset analysis defining adherence as ≥80% was then performed. One hundred one (90.2%) patients were defined as adherent (Table 4), and 11 (9.8%) were nonadherent. All nonadherent participants were on single therapy and prescribed only oral medications. Multivariate regression analysis was limited by the small sample size, so univariate regression analysis was performed (Table 5). A diagnosis of CD was again the only significant predictor for medication adherence (OR = 4.15; 95% CI, 1.09-15.9; P = 0.04).
MARS-5
The mean MARS-5 score at study baseline was 22.3 (SD = 2.54), signaling high adherence, and it was 22.7 (SD = 2.50) at the 52-week point (Table 2). Test-retest over 1 year for the MARS-5 indicated moderate correlation (r = 0.57); similarly, test-retest of medication monitoring adherence at week 0 and week 52 was moderately correlated (r = 0.59). When comparing the MARS-5 measure to the medication monitoring adherence data, there was a moderate correlation between the 2 measures at week 0 (r = 0.48) and week 52 (r = 0.57).
An analysis to determine an appropriate cutoff point for medication adherence on the MARS-5 scale was performed for adherence levels ≥80% and ≥90% (Table 6). At adherence ≥80%, a MARS-5 value of >15 had the highest sensitivity (100%) and lowest specificity (0%) and a MARS-5 value of >25 showed the highest specificity (100%) and lowest sensitivity (9.09%). Sensitivity and specificity were maximized at a value of >22 (sensitivity = 82.5%, specificity = 81.2%), suggesting this value as the best cutoff. At adherence ≥90%, a MARS-5 value of >15 had the highest sensitivity (100%) and lowest specificity (5.0%) and a MARS-5 value of >25 had the highest specificity (100%) and lowest sensitivity (0%). Sensitivity and specificity were maximized at a value of >23 (sensitivity = 70.5%, specificity = 75.0%).
TABLE 6.
MARS-5 as Compared to Medication Monitoring Adherence ≥80% and ≥90%
| MARS-5 | Adherence ≥80%* | Adherence ≥90%† | ||
|---|---|---|---|---|
| Sensitivity | Specificity | Sensitivity | Specificity | |
| >15 | 100% | 9.1% | 100% | 5.0% |
| >16 | 100% | 18.2% | 100% | 10.0% |
| >17 | 99.0% | 18.2% | 98.9% | 10.0% |
| >18 | 97.9% | 27.3% | 98.9% | 20.0% |
| >19 | 95.9% | 27.3% | 96.6% | 20.0% |
| >20 | 92.8% | 54.6% | 93.2% | 35.0% |
| >21 | 87.6% | 54.6% | 87.5% | 35.0% |
| >22‡ | 82.5%‡ | 81.8%‡ | 85.2% | 65.0% |
| >23§ | 69.1% | 100% | 70.5%§ | 75.0%§ |
| >24 | 55.7% | 100% | 59.1% | 90.0% |
| >25 | 35.1% | 100% | 26.4% | 90.0% |
*C-statistic = 0.898.
†C-statistic = 0.794.
‡Numbers represent optimal cutoff point for adherence ≥80%.
§Numbers represent optimal cutoff point for adherence ≥90%.
Logistic regression analysis was then performed on the MARS-5 validation to determine the difference in adherence for each additional MARS-5 point in the cohort. For each additional MARS-5 point, an individual has a 1.7 times greater odds of having adherence ≥90% (95% CI, 1.34-2.24; P < 0.001). For example, a person with a MARS-5 of 21 will have a 1.74 = 8.96 times greater odds of ≥90% adherence than a person with a MARS-5 of 17 (because 21 is 4 points higher than 17, so we raise the OR to the power of 4). Fig. 1 further illustrates that a cutoff of 22 was associated with high adherence as reflected in the biweekly medication monitoring data.
FIGURE 1.
Scatterplot of mean adherence and mean MARS-5 scores.
DISCUSSION
This study is the first to analyze the utility of the MARS-5 as an assessment tool for medication adherence in an IBD population. Previous studies examining medication adherence in IBD-specific populations have utilized a variety of adherence measures, including the Morisky Medication Adherence Scale-8, the MARS-4, the MARS-5, self-reported percentages, the Beliefs About Medicine Questionnaire, and the Visual Analogue Scale.2-4, 10, 11, 19, 20 In addition, predictors of medication adherence have been inconsistent across diseases and in IBD-specific studies. Our study reports a highly adherent cohort of patients with IBD, with a diagnosis of CD as the sole significant predictor for medication adherence. We further aimed to validate the MARS-5 as an accurate assessment tool for medication adherence in an IBD population, showing a moderate correlation when compared to medication monitoring adherence data.
Medication Adherence Measurement
Our study found a moderate correlation between the MARS-5 score and adherence based on biweekly medication usage reporting at 2 time points, a year apart, suggesting variability within individuals for assessment measures throughout the year. Our study supports previous data showing that variability among reporting scales used in adherence studies has made determining the true adherence rates in patients with IBD difficult.11, 20, 25 Ediger et al11 were the first to report medication adherence data in an IBD cohort using the MARS-5. In their article, medication adherence was defined as a MARS-5 score ≥20, which was selected based on positive skewing of the sample distribution and in consultation with the scale developer. Furthermore, they chose a score of ≥20 because it was extrapolated to represent 80% adherence (20/25 = 80%). This aligned with the value used in other studies examining adherence in IBD cohorts and with the original MARS-4- that was previously validated by Selinger et al.5, 19, 20 Our sensitivity and specificity analyses with this IBD sample suggest that a value of >22 on the MARS-5 is an appropriate cutoff to reflect an adherence value of ≥80%.
Studies in other diseases have shown inconsistency among the validity of the MARS-5 and adherence rates, emphasizing the importance of validating adherence for specific disease types. Tommelein et al15 examined the correlation between MARS-5 and self-reported medication adherence in a population with chronic obstructive pulmonary disease. Those authors, who also had a highly adherent population (mean MARS-5 = 23.49; SD = 2.60; mean self-reported adherence = 83.4%), reported a very poor correlation between the 2 measures (Pearson correlation coefficient = 0.103, P = 0.011). In the MARS-4 validation by Selinger et al20 using thiopurine metabolite concentrations in IBD, only a modest correlation was found (r = 0.29, P = 0.005). Furthermore, an Italian study attempted to validate the MARS-5 in an IBD population.26 After validating the Italian language MARS-5I with the English version, those authors compared adherence using physician-evaluated adherence and a Treatment Satisfaction Questionnaire to Medication to the MARS-5I, using the assumption that patients who are more satisfied are more adherent. Although there was no association with the treatment questionnaire, the authors reported a small but statistically significant convergent validity (Spearman’s rho = 0.15; P = 0.014) with the physician-reported adherence measures.26 Our results report a moderate albeit much stronger correlation between biweekly medication monitoring data and the MARS-5 as compared with these studies, suggesting that the MARS-5 can be a simple, valuable assessment tool in measuring medication adherence as compared to an intensive medication monitoring strategy.
Predictors of Adherence
We found that disease type (ie, CD) was the only predictor of medication adherence as assessed by the biweekly medication monitoring. Previous work has reported IBD subtype as a predictor of adherence, but this has been inconsistent. Eindor-Abarbanel et al16 reported UC to be associated with low adherence (OR = 1.748) in a cohort with similar disease distribution as ours (CD = 70.4%), while Ediger et al11reported the same findings but in men only, and they reported immunosuppressant use in women as a predictor of high adherence (using adherence defined as MARS-5 ≥20). Other studies have not reported any difference in disease characteristics.14 Given the small number of our participants who reported adherence <80% and were defined as nonadherers, an adjusted logistic regression analysis was not feasible for this subgroup and thus we performed the adjusted analysis for an adherence of ≥90%.
Because the route of administration has been shown to influence adherence, the higher adherence rates in CD can likely be explained by the increased use of biologic agents in CD.4, 5, 14, 27 Selinger et al5 reported an increase in adherence using more aggressive therapies, with a significant increase between oral agents and biologic therapies, while another study reported lower adherence to topical and oral formulations.14 Only oral medication users reported nonadherence (<80%) in our study, with biologic users all reporting good adherence ≥80% (n = 65). The <90% nonadherence subgroup of 20 patients included only 1 infusion biologic user and 1 adalimumab user, with the remaining being oral medication users (n = 18). The single adalimumab user reported 100% adherence to the subcutaneous injections and 8.6% adherence to the oral medication, therefore skewing the patient’s overall adherence to <90%.
Whereas the majority of our participants on combination therapy reported high adherence to both oral and biologic therapy, it has been previously reported that combination therapy does not change adherence to oral medications.17 In another study, combination therapy was reported to improve medication adherence in persons with either CD or UC.14 Our overall findings support the majority of studies concluding that the route of medication plays a role in medication adherence.
Our study did not show an association between perceived stress and medication adherence. The largest meta-analysis on medication adherence in an IBD population included 10 studies examining psychosocial variables.10 Psychiatric diagnoses and perceived stress were found to be predictors of nonadherence. However, these findings were contradicted in other studies, with 1 finding that perceived stress had a positive association with medication adherence.10 Many psychosocial factors influencing medication adherence have been analyzed in other studies. A review article by Osterberg and Blaschke9 examined barriers to medication adherence, finding that barriers under patient control, including forgetfulness, other priorities, decision to omit doses, and emotional factors are often cited as reasons to not take medications as prescribed. Nonadherence has been reported to be further compounded by psychiatric illness and physician prescribing habits, including complex prescribing regimens and poor explanations on the side effects and rationale for taking a medication.9, 10, 14, 17, 28 One study utilizing the MARS-4 as an adherence measure did not find an association between psychiatric illness and adherence but did report higher adherence with aggressive therapies (ie, biologics) and belief of medication necessity.5 In comparison, Campos et al17 reported depression as a significant predictor of nonadherence in patients with IBD who were prescribed immunomodulators, suggesting variability among studies reporting psychological predictors of adherence. Our study did not examine specific psychiatric diagnoses and their association to adherence in this sample.
STRENGTHS AND LIMITATIONS
The strengths of the study include its prospective design, variety of medication routes prescribed, and frequent tracking of medication use. This is the first large-scale study utilizing regular medication monitoring adherence data in an IBD cohort to validate MARS-5 as an accurate assessment tool.
Our study was limited by a number of factors. It is possible that there was self-selection of more adherent participants, given the nature of the study as a prospective, 1-year study requiring biweekly survey completion. Furthermore, more than 70% of the study participants were women, which has previously been shown to influence survey participation and may also have biased the results for higher adherence.29, 30 Social desirability bias may have been present in our study: patients may have reported higher adherence knowing that their reporting might be seen by a study monitor. Although this bias may have resulted in higher adherence reporting, we do not believe that this would have influenced the association between adherence and other factors in the model. Finally, we were not able to compare adherence to objective measures, such as drug levels, drug metabolite levels, or pharmacy refill data, which is a limitation in any survey-based adherence study. However, regular medication use reporting on a biweekly basis can be reasonably objective to minimize recall bias of missed or changed medications.
A limitation of our study is that not all patients who were otherwise eligible for the study had enough biweekly medication adherence data to compute a medication adherence percentage. To assess the possible impact that those with missing medication adherence may have had on our results, we conducted 2 sensitivity analyses in our logistic regression modeling. In one sensitivity analysis, we assumed that all those with missing data had high adherence (>90%), and in the other sensitivity analysis, we assumed that all those with missing data had low adherence (<90%). Because the logistic regression results were almost identical to those found when the 16 people with missing medication adherence were excluded from analysis, we concluded that there was little evidence that the missing data may have biased our results.
CONCLUSIONS
We report the MARS-5 as a useful medication adherence tool in an IBD population and further describe predictors of adherence in this IBD cohort. We found a moderate correlation between the MARS-5 and biweekly medication monitoring data and satisfactory sensitivity and specificity with the identified cutpoint, suggesting that it is a good tool for assessing adherence in a highly adherent sample such as ours. We report only a diagnosis of CD as a predictor of medication adherence. Predictors of medication adherence in other studies have been inconsistent among the IBD population. Further research needs to be undertaken to determine the best way to assess medication adherence in patients with IBD on long-term maintenance therapy.
Author contributions: James Stone: analysis and interpretation of data, drafting of the manuscript, critical revision of the manuscript for important intellectual content. Leigh Anne Shafer: study concept and design, analysis and interpretation of data, critical revision of the manuscript for important intellectual content, statistical analysis. Lesley Graff: acquisition of funding, study concept and design, analysis and interpretation of data, critical revision of the manuscript for important intellectual content. Lisa Lix: study concept and design, analysis and interpretation of data, critical revision of the manuscript for important intellectual content. Kelcie Witges: study concept and design, acquisition of data, critical revision of the manuscript for important intellectual content, technical or material support. Laura Targownik: study concept and design, acquisition of data, analysis and interpretation of data, critical revision of the manuscript for important intellectual content. Clove Haviva: study concept and design, analysis and interpretation of data, critical revision of the manuscript for important intellectual content. Kathryn Sexton: acquisition of data, analysis and interpretation of data, critical revision of the manuscript for important intellectual content. Charles N. Bernstein: acquisition of funding, study concept and design, acquisition of data, analysis and interpretation of data, critical revision of the manuscript for important intellectual content, technical or material support, study supervision.
Supported by: Canadian Institutes of Health Research.
Conflicts of interest: Dr. Bernstein has consulted for AbbVie Canada, Janssen Canada, Pfizer Canada, Shire Canada, and Takeda Canada, and has received unrestricted educational grants from AbbVie Canada, Janssen Canada, Pfizer Canada, Shire Canada, and Takeda Canada. He has been on the speaker’s bureau of AbbVie Canada, Janssen Canada, Medtronic Canada, and Takeda Canada. He has received a research grant from AbbVie Canada and contract grants from AbbVie, Janssen, Pfizer, Celgene, Roche, and Boeringher Ingelheim. Dr. Targownik has consulted for or been on the advisory boards of Takeda Canada, AbbVie Canada, Ferring Canada, Merck Canada, Pfizer Canada, and Janssen Canada. She has received research grant support from Pfizer Canada. She has been on the speaker’s bureaus for Janssen Canada, Takeda Canada, and Pfizer Canada. The other authors have no conflicts of interest to declare.
Guarantor of the article: Charles Bernstein is the guarantor of the article.
REFERENCES
- 1. Perry J, Chen A, Kariyawasam V, et al. Medication non-adherence in inflammatory bowel diseases is associated with disability. Intest Res. 2018;16:571–578. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Perry J, Chen A, Collins G, et al. Medication non-adherence in inflammatory bowel diseases is significantly associated with disability. J Gastroenterol Hepatol. 2015;30:135–136. [Google Scholar]
- 3. Kane S, Huo D, Aikens J, et al. Medication nonadherence and the outcomes of patients with quiescent ulcerative colitis. Am J Med. 2003;114:39–43. [DOI] [PubMed] [Google Scholar]
- 4. Chan W, Chen A, Tiao D, et al. Medication adherence in inflammatory bowel disease. Intest Res. 2017;15:434–445. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Selinger CP, Eaden J, Jones DB, et al. Modifiable factors associated with nonadherence to maintenance medication for inflammatory bowel disease. Inflamm Bowel Dis. 2013;19:2199–2206. [DOI] [PubMed] [Google Scholar]
- 6. Higgins PD, Rubin DT, Kaulback K, et al. Systematic review: impact of non-adherence to 5-aminosalicylic acid products on the frequency and cost of ulcerative colitis flares. Aliment Pharmacol Ther. 2009;29:247–257. [DOI] [PubMed] [Google Scholar]
- 7. Torres J, Boyapati RK, Kennedy NA, et al. Immunomodulators or biologic agents from patients. Gastroenterology. 2015;149:1716–17 30. [DOI] [PubMed] [Google Scholar]
- 8. Qiu Y, Chen BL, Mao R, et al. Systematic review with meta-analysis: loss of response and requirement of anti-TNFα dose intensification in Crohn’s disease. J Gastroenterol. 2017;52:535–554. [DOI] [PubMed] [Google Scholar]
- 9. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353:487–497. [DOI] [PubMed] [Google Scholar]
- 10. Jackson CA, Clatworthy J, Robinson A, et al. Factors associated with non-adherence to oral medication for inflammatory bowel disease: a systematic review. Am J Gastroenterol. 2010;105:525–539. [DOI] [PubMed] [Google Scholar]
- 11. Ediger JP, Walker JR, Graff L, et al. Predictors of medication adherence in inflammatory bowel disease. Am J Gastroenterol. 2007;102:1417–1426. [DOI] [PubMed] [Google Scholar]
- 12. Sjölander M, Eriksson M, Glader EL. The association between patients’ beliefs about medicines and adherence to drug treatment after stroke: a cross-sectional questionnaire survey. BMJ Open. 2013;3:e003551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Norberg S, Gustafsson M. Older peoples’ adherence and awareness of changes in drug therapy after discharge from hospital. Pharmacy (Basel). 2018;6:1–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Severs M, Mangen MJ, Fidder HH, et al. Clinical predictors of future nonadherence in inflammatory bowel disease. Inflamm Bowel Dis. 2017;23:1568–1576. [DOI] [PubMed] [Google Scholar]
- 15. Tommelein E, Mehuys E, Van Tongelen I, et al. Accuracy of the Medication Adherence Report Scale (MARS-5) as a quantitative measure of adherence to inhalation medication in patients with COPD. Ann Pharmacother. 2014;48:589–595. [DOI] [PubMed] [Google Scholar]
- 16. Eindor-Abarbanel A, Naftali T, Ruhimovich N, et al. Revealing the puzzle of nonadherence in IBD—assembling the pieces. Inflamm Bowel Dis. 2018;24:1352–1360. [DOI] [PubMed] [Google Scholar]
- 17. Campos S, Portela F, Sousa P, et al. Inflammatory bowel disease: adherence to immunomodulators in a biological therapy era. Eur J Gastroenterol Hepatol. 2016;28:1313–1319. [DOI] [PubMed] [Google Scholar]
- 18. Trindade AJ, Ehrlich A, Kornbluth A, et al. Are your patients taking their medicine? Validation of a new adherence scale in patients with inflammatory bowel disease and comparison with physician perception of adherence. Inflamm Bowel Dis. 2011;17:599–604. [DOI] [PubMed] [Google Scholar]
- 19. Jeganathan J, Lee CH, Rahme A, et al. Pediatric-to-adult transition and medication adherence in patients with inflammatory bowel disease. Inflamm Bowel Dis. 2017;23:1065–1070. [DOI] [PubMed] [Google Scholar]
- 20. Selinger CP, Ochieng AO, George V, et al. The accuracy of adherence self-report scales in patients on thiopurines for inflammatory bowel disease: a comparison with drug metabolite levels and medication possession ratios. Inflamm Bowel Dis. 2019;25:919–924. [DOI] [PubMed] [Google Scholar]
- 21. Lee CS, Tan JHM, Sankari U, et al. Assessing oral medication adherence among patients with type 2 diabetes mellitus treated with polytherapy in a developed Asian community: a cross-sectional study. BMJ Open. 2017;7:e016317. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Garfield S, Clifford S, Eliasson L, et al. Suitability of measures of self-reported medication adherence for routine clinical use: a systematic review. BMC Med Res Methodol. 2011;11:1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Witges K, Targownik LE, Haviva C, et al. Living with inflammatory bowel disease: protocol for a longitudinal study of factors associated with symptom exacerbations. JMIR Res Protoc. 2018;7:e11317. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Sexton KA, Walker JR, Targownik LE, et al. The inflammatory bowel disease symptom inventory: a patient-report scale for research and clinical application. Inflamm Bowel Dis. 2019;25:1277–1290. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Severs M, Zuithoff PN, Mangen MJ, et al. Assessing self-reported medication adherence in inflammatory bowel disease: a comparison of tools. Inflamm Bowel Dis. 2016;22:2158–2164. [DOI] [PubMed] [Google Scholar]
- 26. Scribano ML, Caprioli F, Michielan A, et al. Translation and initial validation of the Medication Adherence Report Scale (MARS) in Italian patients with Crohn’s disease. Dig Liver Dis. 2019;51:640–647. [DOI] [PubMed] [Google Scholar]
- 27. Horne R, Parham R, Driscoll R, et al. Patients’ attitudes to medicines and adherence to maintenance treatment in inflammatory bowel disease. Inflamm Bowel Dis. 2009;15:837–844. [DOI] [PubMed] [Google Scholar]
- 28. Horne R, Weinman J. Patients’ beliefs about prescribed medicines and their role in adherence to treatment in chronic physical illness. J Psychosom Res. 1999;47:555–567. [DOI] [PubMed] [Google Scholar]
- 29. Bista K Examining factors impacting online survey response rates in educational research: perceptions of graduate students. J Multidiscip Eval. 2017;13:63–74. [Google Scholar]
- 30. Smith G Does gender influence online survey participation?: A record-linkage analysis of university faculty online survey response behavior ERIC Document Reproduction Service No. ED 501717. 2008.