Association of Physician-Patient Communication and Treatment Satisfaction with Adherence to Disease-Modifying Therapies Among Patients with Relapsing-Remitting Multiple Sclerosis: A Cross-Sectional Study

Abstract

Background

Adherence to disease modifying therapies (DMTs) is essential for treatment outcomes in relapsing-remitting multiple sclerosis (RRMS). Non-adherence has been associated with an increased risk of relapses, disease progression and higher healthcare costs; therefore, it is essential to identify factors associated with adherence.

Purpose

The aim of the study was to evaluate self-reported adherence to therapeutic recommendation among patients with RRMS and to examine its associations with treatment satisfaction and physician-patient communication.

Material and Methods

This cross-sectional online survey study included 165 patients with RRMS (67.27% women), with mean age 44.42 ± 10.43 years, treated at the Department of Neurology, Wroclaw Medical University. Adherence was assessed using the Adherence in Chronic Diseases Scale (ACDS), physician-patient communication with the 14-item Polish version of the Communication Assessment Tool (CAT), and treatment satisfaction with the SATMED-Q questionnaire. Associations with the ACDS score were examined by univariable linear regression and by a multivariable model specified a priori on conceptual and clinical grounds (CAT, SATMED-Q, age, sex, EDSS, comorbidities, place of residence, route of administration and disease duration), supported by regression diagnostics and by ordinal and robust-regression sensitivity analyses.

Results

The mean ACDS score was 25.49 ± 2.51 (range 14–28). Applying the established cut-offs, 8 participants (4.8%) were classified as showing low, 80 (48.5%) intermediate and 77 (46.7%) high adherence. Patients reported high satisfaction with physician communication (mean CAT score: 61.61 ± 9.73) and with treatment (mean total SATMED-Q score 78.67 ± 15.91). In the a-priori multivariable model (n = 165; R² = 0.226, adjusted R² = 0.171, overall F p < 0.001), higher CAT scores (adjusted β = 0.058 per point, 95% CI 0.019 to 0.098, p = 0.004) and higher total SATMED-Q scores (adjusted β = 0.031, 95% CI 0.005 to 0.057, p = 0.018) were associated with higher self-reported adherence, whereas a higher EDSS score was associated with lower adherence (adjusted β = −0.552, 95% CI −0.857 to −0.247, p < 0.001). All three associations were of modest magnitude and were confirmed in ordinal and robust-regression sensitivity analyses (overall F(11,153) = 4.07, p < 0.001).

Conclusion

In this single-centre cross-sectional sample of patients with RRMS, higher self-reported treatment satisfaction and more favourable ratings of physician-patient communication were associated with modestly higher self-reported adherence scores, while greater disability was associated with lower adherence. Longitudinal and intervention studies are required to establish directionality and to determine whether communication-focused or shared decision-making interventions can improve adherence to disease-modifying therapies.

Introduction

Multiple sclerosis (MS) is a long-lasting, polyphasic disease of the central nervous system, affecting approximately 2.8 million people worldwide. It is characterized by multifocal injury to the brain and spinal cord, resulting in a variety of symptoms and disability accumulating throughout the disease.¹ The most prevalent form of MS is relapsing remitting one (RRMS), with episodes of exacerbations (new or recurrent symptoms of neurological deficit) interspersed by remissions (periods of stable condition). The current comprehensive approach to management of MS includes disease-modifying treatment (DMTs), symptomatic therapy, rehabilitation, and lifestyle factors modification. The primary goal of DMTs is to alter the natural course of the disease: by reducing the frequency and severity of relapses and slowing progression of disability. During recent years, DMTs have significantly evolved, offering a range of options to manage the disease. They can be categorized according to their impact upon disease outcomes into high-efficacy treatment agents (HET) and moderate-efficacy therapies (MET).²

In addition, they differ in their mode of administration and may be delivered orally, via subcutaneous or intramuscular injection or intravenously. A choice of optimal treatment should therefore be individualized, considering clinical characteristics, disease activity, risk factors, safety profiles and patient preferences. However, predicting treatment response remains challenging and several non-biological factors may influence long-term treatment outcomes.^3,4

Among these factors, adherence to the therapy plays a crucial role, particularly in the long-term management of MS. Non-adherence to DMTs has been associated with an increased risk of relapses, faster disease progression, higher rates of hospitalizations and greater healthcare and societal costs.^3–5 Therefore, it seems essential to understand and modify factors affecting adherence to treatment in MS patients. In addition to disease-related and treatment-related factors, patient-related aspects-including individual beliefs, expectations and psychosocial characteristics may substantially influence adherence.⁶

Two related but distinct constructs are often conflated in this field and should be separated explicitly. Adherence (implementation) denotes the extent to which a patient’s actual dosing corresponds to the prescribed regimen over the period during which the medicine is being taken, whereas persistence denotes the length of time between initiation and eventual discontinuation of therapy.⁷ The two are driven by partly different mechanisms and are captured by different instruments. The present study addresses adherence in the implementation sense, assessed with a self-report scale; while persistence was not measured.

There are multiple potential contributors to adherence to DMTs. Beyond the beliefs and expectations noted above, it may be shaped by the degree of physical disability and mobility, cognitive functioning, depression and other psychiatric comorbidities, the occurrence and tolerability of adverse effects, regimen complexity and dosing frequency, health literacy, the extent to which patients are involved in shared decision-making, and the accessibility of specialist healthcare services.^6,8,9 No single survey-based study can capture this entire set, and variables that remain unmeasured represent potential residual confounding. However, it is desirable to undertake the assessment of particular factors affecting adherence and potential links between them.

The organisation of MS care forms an important part of this context. In Poland, DMTs are dispensed exclusively within nationally regulated therapeutic programmes: eligibility criteria, the choice of drug, the schedule of monitoring visits and the rules for treatment switching are defined centrally, treatment is free of charge to the patient, and initiation and follow-up are carried out by designated specialist centres according to a unified model. Financial barriers to treatment and variation in prescribing practice, both of which are prominent determinants of adherence in other healthcare systems, are therefore largely removed, while regular attendance at monitoring visits is a condition of continued treatment. This setting is directly relevant to the concept of the study and interpretation of findings. Evidence from central and eastern European MS populations treated within publicly funded therapeutic programmes is scarce. Thus investigation conducted in the present cohort, recruited from a single Polish specialist centre, allows to address various aspects of adherence, without considering cost-sharing issues as potential limitation to treatment availability.

Several previous studies have separately examined treatment satisfaction, physician-patient communication, shared decision-making and adherence in MS.^10–15 What remains insufficiently understood, is how patients’ evaluation of the communication relates to adherence (simultaneously taking into account demographics, clinical characteristics and treatment satisfaction), and how such association is reflected on the scale of the adherence measure itself. These questions certainly deserve thorough investigation.

The aim of this cross-sectional study was to evaluate self-reported adherence to DMTs in patients with RRMS and to examine its possible associations with treatment satisfaction and physician-patient communication. Given the cross-sectional design, the study was designed to quantify associations measured at a single time point rather than to establish temporality, to predict future adherence, or to demonstrate that communication or treatment satisfaction causes better adherence.

Materials and Methods

Study Design and Participants

A cross-sectional online survey-based study was conducted in the patients with relapsing-remitting multiple sclerosis (RRMS), who were under regular follow-up at Department of Neurology, Wroclaw Medical University. “Regular follow-up” was operationalised as attendance at the monitoring visits mandated by the national therapeutic programme for the drug concerned; patients who had discontinued DMT or had been lost to follow-up were consequently not eligible. The study comprised adults (aged over 18 years), diagnosed with RRMS according to McDonald’s criteria (version 2017), who had been treated with DMTs for at least 6 months. Those who were not willing to participate or those who had severe cognitive impairment impeding response to the self-assessment questionnaires, were excluded from the study. Severe cognitive impairment was identified by the treating neurologist based on the clinical record and routine clinical judgement, rather than by a formal cognitive test with standardized cut-off values. Furthermore, patients receiving intravenous infusion therapies (eg natalizumab, ocrelizumab) were excluded from the analysis, because such treatment is administered exclusively on-site in a hospital infusion suite, so it cannot be evaluated in terms of adherence to the self-administered oral or injectable therapeutic agents.

All patients who met the eligibility criteria and were under active follow-up at the centre during the recruitment period were invited consecutively, so that the invited group corresponded to the whole eligible clinic population rather than to a selected subgroup. Of the 220 eligible patients invited, 170 submitted a complete response, giving a response rate of 77%. The electronic form required every item to be answered before the questionnaire could be submitted, so no partial responses were received and no participant contributed incomplete data. Of these 170 respondents, five receiving intravenous infusion therapies were excluded, giving the analysis sample of 165.

Prior to participation, individuals reviewed and signed an informed consent form explaining the study’s aims, associated risks and benefits, and the safeguards used to protect personal data. Consent was obtained electronically: the information sheet was displayed as the first page of the survey and participants could proceed to the questionnaire only after actively confirming their consent; no data were recorded for individuals who did not consent. The survey link was distributed among these patients between September and December 2024. The questionnaire was built and administered as an electronic form using Google Forms (Google LLC, Mountain View, CA, USA). Responses were automatically transferred to an electronic spreadsheet, minimizing the risk of data entry errors. All data were anonymized and coded prior to analysis. Ethical approval was obtained from the Bioethics Committee of Wroclaw Medical University (approval no. KB 175/2022). The study was conducted in accordance with the ethical principles of the Declaration of Helsinki.

Instruments

A questionnaire designed by the authors was used to collect sociodemographic data (age, gender, place of residence, level of education, employment status, marital status) and clinical data including time since MS diagnosis, frequency of relapses, type of treatment (oral vs injectable) and its duration, the specific DMT received, the accompanying symptoms reported by the patient, and the Expanded Disability Status Scale (EDSS) score. The EDSS was determined by the treating neurologist during the patient’s recent scheduled visit, in accordance with the routine monitoring required by the programme, and the score was then provided to the patient within brief summary of the visit outcomes, so that it could be entered in the corresponding field of the electronic questionnaire. Disability is therefore the one variable in the multivariable model that originates in a clinician’s assessment rather than in the patient’s own judgement, although it reaches the dataset through the patient’s transcription and was not additionally verified against the medical record.

The DMTs received by the analysed participants were dimethyl fumarate (Tecfidera, n = 63; oral, twice daily), fingolimod (Gilenya, n = 16; oral, once daily), teriflunomide (Aubagio, n = 14; oral, once daily), peginterferon beta-1a (Plegridy, n = 16; subcutaneous, every 2 weeks), interferon beta-1a subcutaneous (Rebif, n = 15; three times weekly), interferon beta-1a intramuscular (Avonex, n = 15; once weekly), interferon beta-1b (Betaferon/Betaseron, n = 13; subcutaneous, every other day) and glatiramer acetate (Copaxone, n = 13; subcutaneous, daily or three times weekly). For the route-of-administration comparison, the three oral agents were grouped as “oral” (n = 93) and the five injectable agents as “injectable” (n = 72); every participant in the injectable group self-administers a subcutaneous or intramuscular injection. Information on recent treatment switching or temporary treatment interruptions was not collected, and this is acknowledged as a limitation.

Three standardized questionnaires were used:

The Adherence Chronic Diseases Scale (ACDS) used to measure the level of adherence to treatment. The questionnaire comprises seven questions with responses rated on a five-point (A-E) scale scored 4 to 0, giving a total score of 0–28 points. The total score >26 points indicates high adherence; score between 21 and 26 - intermediate adherence and scores below 21 points - low adherence.¹⁶ In the Polish version of the ACDS questionnaire, the value of the Cronbach’s α coefficient was 0.739, which confirmed reliability and homogeneity of the questionnaire.¹⁷ The ACDS is a generic instrument developed for chronic conditions requiring long-term pharmacotherapy rather than an MS-specific tool. It was selected because it is validated in Polish, is not tied to a particular route of administration (therefore permits oral and injectable DMTs to be assessed on the same scale), and because it captures the implementation phase of adherence, which is the construct of interest here; MS-specific alternatives validated in Polish were not available. In the present sample the internal consistency of the ACDS was lower than in the original validation study (Cronbach’s α = 0.562, standardised α = 0.588).

Polish version of the Treatment Satisfaction with Medicines Questionnaire (SATMED-Q), was used to evaluate the satisfaction with the current treatment. It consists of 17 items grouped into six domains: efficacy of treatment (3 items), convenience of use (3 items), impact on daily activities (3 items), medical care (2 items), overall satisfaction (3 items), undesirable side effects (3 items), the last of which is reported in under the label “lack of adverse events” because it is reverse-scored (so that a higher value denotes less burden). Responses in each domain are scored on a five-point Likert scale (0–4). The total sum of scores ranges from 0 to 68 points, with higher score corresponding with greater satisfaction with the therapy. The composite score is then calculated using 0–100 normalized scale.^18,19 Normalisation follows Ruiz et al:¹⁸ Y’ = [(Y - Ymin) / (Ymax - Ymin)] x 100, applied to the raw total (Ymax = 68) and separately to each domain using that domain’s own maximum (12 for the five three-item domains, 8 for medical care). The three items of the undesirable side effects domain are negatively worded and were reverse-scored (4 - raw) before summation, so that a higher score denotes less burden from side effects, consistent in direction with all other domains. No skip rule tied to the screening question on the occurrence of side effects could be verified in the published literature; items 1–3 were therefore scored exactly as answered by every participant. In the present sample Cronbach’s α was 0.898 for the total scale and ranged from 0.867 to 0.947 across domains.

Communication Assessment Tool (CAT) was used to assess patients’ perception of communication with their physicians. The instrument administered was the 14-item Polish short form validated by Swiatoniowska-Lonc et al,²⁰ which omits the item on the treatment of the patient by the physician’s office staff that is present in the original 15-item version.²¹ It includes 14 items rated on a 5-point Likert scale, with higher scores indicating better communication (total score range 14–70).²¹ In the Polish version of the CAT questionnaire, the value of the Cronbach’s α coefficient was 0.96, which indicates very good psychometric properties and high internal consistency of the tool.²⁰ In the present sample Cronbach’s α was 0.979.

The survey additionally contained two single-item anchors that are not part of any of the three instruments: perceived ease of taking the medication as prescribed, and overall satisfaction with treatment during the preceding four weeks. These were not included in any scale score and are reported only as convergent-validity evidence.

Statistical Analysis

Results of the questionnaires were calculated for the study group. Relationships were analyzed between ACDS scores and sociodemographic and clinical data, as well as SATMED-Q and CAT results. Quantitative data were analyzed using descriptive statistics (mean, standard deviation, median, quartiles, and range). Qualitative variables were summarized with absolute and relative frequencies. The distribution of the ACDS score was additionally summarised categorically using the published cut-offs (<21, 21–26, >26). Group comparisons for qualitative variables used the chi-square test (with Yates’ correction for 2x2 tables) or Fisher’s exact test when assumptions were not met. Quantitative variables between the two groups were compared using the Mann–Whitney U-test, across more than two groups using the Kruskal–Wallis test, and associations between ordinal single items and scale scores using Spearman’s rank correlation.

Factors associated with the ACDS total score were examined by linear regression. Each candidate variable was first examined in a univariable model. The multivariable model was then specified a priori on conceptual and clinical grounds rather than by statistical screening, because selection on univariable p-values can exclude clinically important confounders; it includes the two exposures of interest (the CAT and the total SATMED-Q score) together with age, sex, EDSS, comorbidities, place of residence, route of administration and time since diagnosis, irrespective of their univariable significance. Because the dataset is complete, every model is estimated on all 165 participants, and no observation is lost to listwise deletion.

For every model we report the analysed n, the regression coefficient with its 95% confidence interval and exact p-value, R², adjusted R² and the overall model F-test; standardised coefficients are also reported. For categorical predictors with more than two levels, an overall (omnibus) F-test of the whole variable is reported in addition to the individual level contrasts, and pairwise contrasts are adjusted for multiple comparisons by the Holm procedure.²² Multicollinearity was assessed with the variance inflation factor (generalised VIF for multi-level factors).

The assumptions of linear regression were examined for the multivariable model: normality of residuals by the Shapiro–Wilk test and a normal quantile-quantile plot, homoscedasticity by the Breusch-Pagan test and a residual-versus-fitted plot, and influential observations by Cook’s distance (cut-off 4/n). Because the ACDS is a bounded, discrete score with a pronounced ceiling, two pre-specified sensitivity analyses were performed: a proportional-odds ordinal regression on the three established adherence categories, and an MM-type robust linear regression, both with the same predictors as the multivariable model.²³

No a priori sample-size calculation was performed before recruitment, since all eligible patients under follow-up at the centre were invited. The available sample was therefore evaluated against the rule of Green,²⁴ which requires N ≥ 104 + m for testing an individual predictor and N ≥ 50 + 8m for testing the overall R², where m is the number of model terms. With m = 11 terms in the multivariable model and the complete sample of n = 165, both requirements are satisfied (n ≥ 115 and n ≥ 138 respectively).

Results are reported as regression coefficients with 95% confidence intervals. A p-value < 0.05 was considered statistically significant. All statistical analyses were performed using R Software, Version 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria), with the packages psych (internal consistency), car (variance inflation factors, Breusch-Pagan test), MASS (ordinal and robust regression) and emmeans (omnibus tests and Holm-adjusted contrasts). Because several models were estimated, and because no adjustment was made across the full set of exploratory comparisons, all findings other than those concerning CAT, SATMED-Q and EDSS should be regarded as hypothesis-generating.

Results

Sociodemographic and Clinical Characteristics of the Study Group

A total of 165 patients with RRMS (67.27% of women), with mean age 44.42 ± 10.43 were included. Patients receiving injectable drugs were significantly older than those taking tablets (46.75 ± 10.27 vs 42.61 ± 10.24 years; p = 0.013). Most respondents lived in urban areas, nearly half of them had higher education and more than 60% were married and professionally active. Sociodemographic characteristics of the study group are presented in Table 1.

Table 1.

Sociodemographic Characteristics of the Study Group (N = 165)

Variables	n (%) or M ± SD
Gender
Women	111 (67.27%)
Men	54 (32.73%)
Age (years)	44.42 ± 10.43
Place of residence
Village	41 (24.85%)
CITY < 50,000 residents	43 (26.06%)
CITY 50,000-500000 residents	26 (15.76%)
CITY > 500,000 residents	55 (33.33%)
Educational level
Basic	3 (1.82%)
Vocational	24 (14.55%)
Middle	61 (36.97%)
High	77 (46.67%)
Marital status
Single	34 (20.61%)
Married	111 (67.27%)
Divorced	17 (10.30%)
Widow/er	3 (1.82%)
Employment status
Active/working	113 (68.48%)
Retired	16 (9.70%)
Pensioner	24 (14.55%)
Unemployed	10 (6.06%)
Student	2 (1.21%)

Abbreviations: n, number of patients; %, percent; M, mean; SD, standard deviation.

Table 2 summarizes clinical characteristics of the study group. Ninety-three patients took pills, and the rest of the group was treated with injectable drugs. The mean time since MS diagnosis was 12.59 ± 6.76 years and the median EDSS score was 2.0 (Q1-Q3 1.5–3.0; range 1.0–6.5; mean 2.35 ± 1.22), indicating a cohort with mild to moderate disability. Patients receiving injectable treatments had a significantly longer disease duration than those on oral therapies (14.71 ± 7.14 vs 10.95 ± 5.99 years; p < 0.001) and had been treated for longer time (7.44 ± 3.85 vs 6.18 ± 4.03 years; p = 0.033). The two groups did not differ in EDSS (p = 0.255) or in the number of relapses in the previous two years (p = 0.350). The most frequently reported bothersome disease symptoms were fatigue (122 patients, 73.9%), gait and balance disturbance (101, 61.2%), visual disturbance (84, 50.9%) and perceptual or emotional disturbance including low mood (53, 32.1%); only two participants (1.2%) reported no bothersome symptoms.

Table 2.

Clinical Characteristics of the Study Group (N = 165)

Variables	n (%) or M ± SD
Form of medication
Oral	93 (56.36%)
Injectable	72 (43.64%)
Disease-modifying therapy
Dimethyl fumarate, oral, twice daily	63 (38.18%)
Fingolimod, oral, once daily	16 (9.70%)
Teriflunomide, oral, once daily	14 (8.48%)
Peginterferon beta-1a, s.c., every 2 weeks	16 (9.70%)I
Interferon beta-1a, s.c., 3 times weekly	15 (9.09%)
Interferon beta-1a, i.m., once weekly	15 (9.09%)
Interferon beta-1b, s.c., every other day	13 (7.88%)
Glatiramer acetate, s.c., daily or 3 times weekly	13 (7.88%)
EDSS score, M ± SD	2.35 ± 1.22
EDSS score, Me (Q1-Q3); range	2.0 (1.5–3.0); 1.0–6.5
Time since diagnosis (years)	12.59 ± 6.76
Number of relapses during the last 2 years	0.61 ± 1.04
Duration of treatment (years)	6.73 ± 3.99
Comorbidities	62 (37.58%)
Accompanying or bothering symptoms reported (multiple answers possible)
Fatigue	122 (73.94%)
Gait and balance disturbance	101 (61.21%)
Visual disturbance	84 (50.91%)
Sphincter dysfunction (urinary incontinence, urgency, constipation)	56 (33.94%)
Perceptual and emotional disturbance (emotional lability, low mood, depression)	53 (32.12%)
Sexual dysfunction	40 (24.24%)
Increased muscle tone (spasticity)	38 (23.03%)
Sensory disturbance (paraesthesia, neuralgia)	37 (22.42%)
Speech disturbance (scanning speech)	21 (12.73%)
Swallowing disturbance (dysphagia)	13 (7.88%)
No symptoms reported	2 (1.21%)

Notes: All variables have a complete set of 165 responses. Percentages for accompanying symptoms sum to more than 100% because respondents could select more than one symptom.

Abbreviations: n, number of patients; %, percent; M, mean; SD, standard deviation; Me, median; Q1, first quartile; Q3, third quartile; EDSS, Expanded Disability Status Scale; s.c., subcutaneous; i.m., intramuscular.

Adherence to Treatment

Table 3 presents the ACDS and CAT scores. The mean ACDS score was 25.49 ± 2.51 (median 26, Q1-Q3 24–27, range 14–28). Because the mean and the median both fall inside the intermediate band defined by the instrument’s own cut-offs, the categorical distribution is the more informative summary and is given in Table 4: 8 participants (4.8%) scored below 21 (low adherence), 80 (48.5%) scored 21–26 (intermediate adherence) and 77 (46.7%) scored above 26 (high adherence). Adherence in this cohort is therefore appropriately described as intermediate-to-high rather than uniformly high. A pronounced ceiling effect was present: 37 participants (22.4%) obtained the maximum score of 28.

Table 3.

Results of the CAT and ACDS in the Study Group (N = 165)

	n	Mean	SD	Median	Min	Max	Q1	Q3
ACDS	165	25.49	2.51	26	14	28	24	27
CAT	165	61.61	9.73	65	28	70	56	70

Abbreviations: SD, standard deviation; Q1, the first quartile; Q3, the third quartile; ACDS, the Adherence in Chronic Diseases Scale (score range 0–28); CAT, Communication Assessment Tool (14-item Polish short form, score range 14–70).

Table 4.

Distribution of Adherence Categories According to the ACDS Cut-Off Scores (N = 165)

Adherence Category	ACDS Score	n	%
Low adherence	< 21	8	4.85
Intermediate adherence	21–26	80	48.48
High adherence	> 26	77	46.67
Total	0–28	165	100.00

Note: Cut-offs as defined by Kubica et al¹⁶.

Abbreviations: ACDS, the Adherence in Chronic Diseases Scale (score range 0–28).

In unadjusted analyses, the ACDS score did not differ significantly across categories of place of residence (Kruskal–Wallis p = 0.422) or by route of administration (median 26 in both groups, p = 0.576), and the distribution of adherence categories was also comparable between oral and injectable treatment (p = 0.904).

Perception of Physician-Patient Communication

The mean CAT score in the study group was 61.61 ± 9.73 (median 65, Q1-Q3 56–70, range 28–70) (Table 3), which indicated that the respondents were satisfied with communication with their physicians. A ceiling effect was again evident, with 57 of 165 participants (34.5%) obtaining the maximum score of 70. CAT scores did not differ significantly by place of residence (Kruskal–Wallis p = 0.517) or by route of administration (p = 0.782); the group mean was lowest among residents of the largest cities (60.96) and highest in cities of 50,000-500000 inhabitants (63.54).

Treatment Satisfaction

The mean total SATMED-Q score (mean 78.67 ± 15.91, median 80.88, Q1-Q3 70.59–91.18) indicates overall favorable satisfaction with treatment. The results are shown in Table 5. The higher satisfaction scores were observed in the domain of overall satisfaction and medical care, whereas lower scores were noted in the domains of treatment effectiveness and impact on daily activities. Patients receiving oral drugs had significantly higher levels of total score SATMED-Q (median 82.35 vs 79.41, p = 0.025) and convenience of use than those taking injection (median 100 vs 75, p < 0.001).

Table 5.

Treatment Satisfaction with Medicines Questionnaire (SATMED-Q) Results in the Study Group (N = 165)

Domain	M	SD	Me	Min	Max	Q1	Q3
Total score SATMED-Q	78.67	15.91	80.88	27.94	100	70.59	91.18
Lack of adverse events	83.13	24.33	100	0	100	75	100
Treatment effectiveness	67.88	25.69	75	0	100	50	83.33
Convenience of use	79.34	25.46	83.33	0	100	75	100
Impact on daily activities	67.78	28.86	75	0	100	50	91.67
Medical care	87.35	17.78	100	25	100	75	100
Overall satisfaction	89.44	16.23	100	0	100	83.33	100

Notes: Every participant contributed a complete set of items to every domain. The three items of the “lack of adverse events” domain are negatively worded and were reverse-scored before normalisation, so that a higher score denotes less burden from side effects.

Abbreviations: M, mean; SD, standard deviation; Me, median; Min, minimum; Max, maximum; Q1, first quartile; Q3, third quartile.

Factors Associated with Adherence

Univariable linear regression (Table 6) showed that higher CAT scores (β = 0.076 per point, 95% CI 0.038 to 0.114, p < 0.001), higher total SATMED-Q scores (β = 0.037, 95% CI 0.014 to 0.061, p = 0.002) and lower EDSS scores (β = −0.543, 95% CI −0.851 to −0.236, p < 0.001) were associated with higher ACDS scores. No sociodemographic variable, and no other clinical or treatment-related variable, reached statistical significance in univariable analysis.

Table 6.

Univariable Linear Regression: Factors Associated with the ACDS Total Score (N = 165)

Variable	Category	β	95% CI	p
CAT (per 1 point)	-	0.076	0.038 to 0.114	< 0.001 *
SATMED-Q total (per 1 point)	-	0.037	0.014 to 0.061	0.002 *
EDSS (per 1 point)	-	−0.543	−0.851 to −0.236	< 0.001 *
Sex	Female	Ref.	-	-
	Male	0.481	−0.341 to 1.304	0.249
Age (per 1 year)	-	−0.012	−0.049 to 0.025	0.522
Place of residence	Village	Ref.	-	-
	City < 50,000	0.697	−0.385 to 1.778	0.205
	City 50,000-500000	1.091	−0.151 to 2.333	0.085
	City > 500,000	0.339	−0.683 to 1.361	0.513
Educational level	Basic or vocational	Ref.	-	-
	Middle	−0.548	−1.692 to 0.597	0.346
	Higher	0.137	−0.971 to 1.244	0.808
Marital status	Single	Ref.	-	-
	Married	−0.348	−1.325 to 0.629	0.483
	Divorced or widowed	−0.085	−1.489 to 1.319	0.905
Employment status	Unemployed	Ref.	-	-
	Working	0.951	−0.689 to 2.592	0.254
	Retired	0.938	−1.067 to 2.942	0.357
	Pensioner	1.500	−0.372 to 3.372	0.115
	Student	2.500	−1.352 to 6.352	0.202
Route of administration	Oral	Ref.	-	-
	Injectable	0.213	−0.567 to 0.994	0.590
Comorbidities	No	Ref.	-	-
	Yes	−0.140	−0.940 to 0.659	0.729
Time since diagnosis (per 1 year)	-	−0.047	−0.104 to 0.010	0.109
Relapses in the last 2 years (per 1)	-	−0.040	−0.414 to 0.334	0.833
Treatment duration (per 1 year)	-	−0.050	−0.147 to 0.047	0.311

Notes: * denotes p < 0.05. Each row block corresponds to a separate univariable model with the ACDS total score as the dependent variable; all models are estimated on the complete sample of 165 participants. Reference categories are stated explicitly for every categorical variable.

Abbreviations: β, unstandardised regression coefficient; CI, confidence interval; ref., reference category.

The a-priori multivariable model (Table 7) explained 22.6% of the variance in the ACDS score (adjusted R² = 0.171; F(11,153) = 4.07, p < 0.001). Three variables were independently associated with the ACDS score: the CAT score (adjusted β = 0.058, 95% CI 0.019 to 0.098, p = 0.004; standardised β = 0.226), the total SATMED-Q score (adjusted β = 0.031, 95% CI 0.005 to 0.057, p = 0.018; standardised β = 0.197) and the EDSS score (adjusted β = −0.552, 95% CI −0.857 to −0.247, p < 0.001; standardised β = −0.268). Age, sex, comorbidities, route of administration and time since diagnosis were not independently associated with adherence.

Table 7.

Multivariable Linear Regression Model for the ACDS Total Score, Specified a Priori (N = 165)

Variable	Category	β	95% CI	p	std. β
CAT (per 1 point)	-	0.058	0.019 to 0.098	0.004 *	0.226
SATMED-Q total (per 1 point)	-	0.031	0.005 to 0.057	0.018 *	0.197
EDSS (per 1 point)	-	−0.552	−0.857 to −0.247	< 0.001 *	−0.268
Age (per 1 year)	-	0.015	−0.024 to 0.054	0.461	0.061
Sex	Female	Ref.	-	-	-
	Male	0.204	−0.568 to 0.977	0.602	0.081
Comorbidities	No	Ref.	-	-	-
	Yes	0.019	−0.729 to 0.766	0.961	0.007
Place of residence	Village	Ref.	-	-	-
	City < 50,000	1.210	0.194 to 2.225	0.020 *	0.481
	City 50,000-500000	0.891	−0.263 to 2.046	0.129	0.355
	City > 500,000	0.549	−0.398 to 1.495	0.254	0.218
Route of administration	Oral	Ref.	-	-	-
	Injectable	0.452	−0.332 to 1.235	0.256	0.180
Time since diagnosis (per 1 year)	-	−0.049	−0.111 to 0.013	0.123	−0.132
Model summary	n	165
	R2	0.226
	Adjusted R2	0.171
	Overall F-test	F(11,153) = 4.07		< 0.001 *

Notes: * denotes p < 0.05. The model was specified a priori on conceptual and clinical grounds, irrespective of the univariable significance of its terms, and is estimated on the complete sample of 165 participants. The omnibus F-test for place of residence was not significant (F(3,153) = 2.00, p = 0.116) and no pairwise contrast survived Holm correction (smallest adjusted p = 0.119), so the individual contrasts for this variable should not be interpreted as established findings.

Abbreviations: β, unstandardised regression coefficient; std. β, standardised regression coefficient; CI, confidence interval; ref., reference category.

The magnitude of these associations is modest in relation to the range of the outcome. A 10-point higher CAT score, approximately one standard deviation, corresponds to an adjusted increase of 0.58 ACDS points, and a 10-point higher SATMED-Q score to an increase of 0.31 ACDS points, on a scale running from 0 to 28 points on which the interquartile range of the observed data spans only 3 points. A 1-point higher EDSS score corresponds to an adjusted decrease of 0.55 ACDS points. None of these differences would by itself move a typical participant from one adherence category to another.

Place of residence entered the model with one nominally significant contrast (residence in a city of fewer than 50,000 inhabitants vs rural residence, p = 0.020); the two remaining city contrasts were not significant (p = 0.129 and p = 0.254). The omnibus F-test for place of residence as a whole was not significant (F(3,153) = 2.00, p = 0.116), and none of the six pairwise contrasts remained significant after Holm correction (smallest adjusted p = 0.119). The association between place of residence and adherence should therefore be regarded as an exploratory observation rather than an established finding.

Variance inflation factors were low for every predictor in the multivariable model (maximum GVIF^(1/2df) = 1.19), so the conceptual overlap between the CAT and the medical care domain of the SATMED-Q, although reflected in a moderate rank correlation (Spearman ρ = 0.51), did not produce problematic multicollinearity.

Model Diagnostics and Sensitivity Analyses

Diagnostics for the multivariable model are summarised in Table 8. The residuals departed significantly from normality (Shapiro–Wilk W = 0.943, p < 0.001) and 10 of 165 observations (6.1%) exceeded the Cook’s distance cut-off of 4/n, whereas the Breusch-Pagan test gave no evidence of heteroscedasticity (p = 0.139). The departure from normality is consistent with the bounded, ceiling-limited distribution of the ACDS score and motivated the pre-specified sensitivity analyses.

Table 8.

Regression Diagnostics and Sensitivity Analyses for the Multivariable Model (N = 165)

Assessment	Result	p
A. Diagnostics
Normality of residuals (Shapiro–Wilk)	W = 0.943	< 0.001
Homoscedasticity (Breusch-Pagan)	BP = 16.06, df = 11	0.139
Influential observations (Cook’s distance > 4/n)	10 of 165 (6.1%)	-
Multicollinearity (maximum GVIF^(1/2df))	1.19	-
Sample size, individual predictor (Green 1991)	Required n ≥ 115; available n = 165 (met)	-
Sample size, overall R2 test (Green 1991)	Required n ≥ 138; available n = 165 (met)	-
B. Sensitivity analysis: proportional-odds ordinal regression on the three ACDS categories
CAT	β = 0.046 (95% CI 0.010 to 0.082)	0.013
SATMED-Q total	β = 0.033 (95% CI 0.009 to 0.058)	0.007
EDSS	β = −0.381 (95% CI −0.669 to −0.102)	0.008
C. Sensitivity analysis: robust (MM-estimator) linear regression
CAT	β = 0.065 (95% CI 0.031 to 0.100)	< 0.001
SATMED-Q total	β = 0.028 (95% CI 0.006 to 0.050)	0.014
EDSS	β = −0.535 (95% CI −0.800 to −0.270)	< 0.001
D. Place of residence: omnibus test and multiple comparisons
Omnibus F-test for the whole variable	F(3,153) = 2.00	0.116
Pairwise contrasts, Holm-adjusted	All 6 contrasts non-significant	≥ 0.119

Abbreviations: β, regression coefficient; CI, confidence interval; GVIF, generalised variance inflation factor; df, degrees of freedom.

Both sensitivity analyses confirmed all three principal findings. In a proportional-odds ordinal regression on the three adherence categories, the CAT score (β = 0.046, 95% CI 0.010 to 0.082, p = 0.013), the total SATMED-Q score (β = 0.033, 95% CI 0.009 to 0.058, p = 0.007) and the EDSS score (β = −0.381, 95% CI −0.669 to −0.102, p = 0.008) retained their direction and significance. In robust regression the same pattern was observed for the CAT score (β = 0.065, p < 0.001), the total SATMED-Q score (β = 0.028, p = 0.014) and the EDSS score (β = −0.535, p < 0.001). The three associations are therefore stable across the ordinary least-squares, ordinal and robust specifications.

Convergent Validity of the Instruments

Two single-item anchors collected in the same survey but not forming part of any scale were used to examine convergent validity (Table 9). Overall satisfaction with treatment during the preceding four weeks correlated significantly with all three instruments, most strongly with the total SATMED-Q score (Spearman ρ = 0.505, p < 0.001) and more weakly but significantly with the CAT (ρ = 0.266, p < 0.001) and the ACDS (ρ = 0.229, p = 0.003). Perceived ease of taking the medication as prescribed correlated with the total SATMED-Q score (ρ = 0.307, p < 0.001) but not with the CAT (p = 0.353) or the ACDS (p = 0.354). The pattern is consistent with the intended content of each instrument, and the absence of a correlation between perceived ease of administration and the ACDS suggests that the adherence score is not simply a restatement of how convenient patients find their regimen.

Table 9.

Convergent Validity: Spearman Correlations Between Two Single-Item Anchors and the Three Study Instruments (N = 165)

Single-Item Anchor	ACDS: ρ (p)	CAT: ρ (p)	SATMED-Q Total: ρ (p)
Perceived ease of taking the medication as prescribed	0.073 (0.354)	0.073 (0.353)	0.307 (< 0.001) *
Overall satisfaction with treatment during the last 4 weeks	0.229 (0.003) *	0.266 (< 0.001) *	0.505 (< 0.001) *

Notes: ρ - Spearman rank correlation coefficient. * denotes p < 0.05. Both anchors are single researcher-added items that are not part of any of the three scored instruments; they were collected in the same survey and are reported here as convergent-validity evidence.

Discussion

Over the past decades, disease-modifying therapies in MS have evolved significantly, offering a broad spectrum of treatment options that differ in mechanism of action, efficacy, safety profiles, and routes of administration. The therapeutic expansion enables a more individualized approach to MS management.⁴ Despite these advances, adherence to DMTs still remains a significant challenge with reported rates ranging widely from 27% up to 90%, depending on the population and methods used to assess adherence.^10,11,25 Importantly, non-adherence has been associated with a higher risk of relapses, progression of disability, poorer quality of life and increased healthcare costs.^6,26

In the present study, adherence assessed using the ACDS scale was intermediate to high among patients with RRMS, with just under half of the cohort reaching the high-adherence category and a further half falling in the intermediate band. This distribution is broadly consistent with previous reports,^5,10,25 although adherence levels reported in the literature vary considerably, reflecting differences in healthcare systems, in the instruments used and patients’ population. Comparisons across studies are complicated further by the frequent conflation of adherence and persistence,⁷ and by the fact that self-report instruments systematically yield higher estimates than pharmacy-refill or electronic-monitoring measures.

Among sociodemographic variables, none showed a robust association with adherence in this cohort. Place of residence produced one nominally significant dummy contrast in the multivariable model, but the association did not survive an omnibus test of the variable as a whole or Holm correction of the pairwise contrasts, and it should therefore be treated as exploratory. Plausible mechanisms would include differences in access to healthcare services and in the availability of specialist MS centres,^8,27 but the present data do not provide evidence strong enough to support such an interpretation. Age was not associated with adherence in either univariable or multivariable analysis. Reports on age in the literature remain ambiguous,^8,28 which is consistent with an effect that is small and context dependent.

The only clinical variable associated with adherence was degree of disability. A higher EDSS score was associated with a lower ACDS score in univariable analysis, in the a-priori multivariable model and in both sensitivity analyses. Greater physical disability may hinder adherence through reduced manual dexterity and mobility, greater dependence on assistance with self-administration, and a higher burden of symptoms, including fatigue and the cognitive decline,⁹ The symptom profile of this cohort is consistent with such a mechanism, with fatigue reported by 73.9% and gait or balance disturbance by 61.2% of participants. Reverse pathways are equally plausible in a cross-sectional design, since poorer long-term adherence may itself contribute to disability accumulation, and the direction of this association cannot be resolved here. No association was observed with disease duration, relapse frequency or route of administration. Some previous reports have linked longer disease duration to poorer adherence,^29,30 which we did not replicate. Regarding route of administration, the literature is divided, with higher adherence reported for oral agents in some studies,^31,32 the opposite in others³³ and no difference in further reports.^28,34

Apart from disease and treatment-related factors, patient-related aspects - particularly subjective perceptions of treatment and the quality of physician-patient communication - play crucial but often underestimated role in adherence. Therefore, we focused on treatment satisfaction and physician-patient communication as a key component of the patient perspective.

Treatment satisfaction is defined as patients’ evaluation of their experience, in relation to expectation and outcomes,³⁵ and can be affected by various aspects of treatment. In the present study SATMED-Q results indicated moderate satisfaction with treatment effectiveness and impact on daily activities, alongside very high satisfaction with medical care. Similarly, CAT scores demonstrated a generally positive perception of physician-patient communication. These results are consistent with previous studies.^36–38

In previous studies conducted in various MS populations, patient-physician communication was rated as moderate to high.^36–38 It is worth highlighting that the very high ratings of medical care and communication observed here should be read in view of the specificity of national therapeutic programme in Poland: all participants were followed regularly within unified regimen, at a single specialist MS centre that has implemented a comprehensive patient-centred model of care. Both the level of the scores and the restricted range they occupy are therefore likely to be specific to this setting.

Evidence from several studies indicates that engaging patients in therapeutic decisions and follow-up of the treatment outcomes improves their overall quality of life but also their perception of treatment.^11,12 Furthermore, adherence to DMT was shown to be significantly associated with treatment satisfaction.^12–15 Specifically, effective communication correlated with better adherence in the management of chronic disorders such as diabetes, depression, hypertension or glaucoma.^39–41 Satisfactory communication between MS patients and physicians has been shown to enhance treatment adherence.^11,12,39

Consistent with these reports, CAT and total SATMED-Q scores in our cohort were independently associated with the ACDS score, and both associations were reproduced in the ordinal and robust sensitivity analyses. One qualification is nevertheless essential: the magnitude of the associations is modest. An increase of one standard deviation in the CAT score corresponds to roughly half a point on a 28-point adherence scale, and the corresponding value for the SATMED-Q is smaller still. Whether differences of this size translate into a clinically meaningful difference in medication-taking behaviour is unknown, and the present data cannot establish it. The full multivariable model accounts for less than a quarter of the variance in adherence (R² = 0.226, adjusted R² = 0.171), which is itself a reminder that most of the variation in adherence in this cohort is explained by factors that were not measured.

Directionality cannot be determined from these data, and reverse causation is entirely plausible: patients who take their medication consistently may experience fewer symptoms and fewer treatment failures and may in consequence report greater satisfaction and a more positive view of their physician. Because adherence, satisfaction and communication were all measured by self-report in a single questionnaire administered at one time point, common-method variance may additionally have inflated the observed associations between them.⁴² This concern does not apply in the same way to the association with disability, since the EDSS was rated by the treating neurologist rather than by the patient, and the EDSS finding is correspondingly less vulnerable to this bias. The results are therefore best read as showing that patients’ perception of care and their self-reported adherence form a coherent cluster, without warranting any statement about impact of one element upon another.

The study has several strengths. The population was well defined, with a confirmed diagnosis and documented treatment status within a nationally regulated therapeutic programme; validated instruments were used throughout; the dataset was verified against the source survey records before analysis, so that every analysis is based on the complete sample of 165 participants with no loss to missing data. The analysis incorporated demographic, clinical and treatment-related variables simultaneously, including a clinician-rated measure of disability, determined by the treating neurologist at the programme visit, that is frequently absent from survey-based adherence studies. The reporting of regression diagnostics and of two pre-specified sensitivity analyses allows the robustness of each finding to be judged directly.

Several limitations must nevertheless be acknowledged. First, the cross-sectional design precludes any inference about temporality or causation. Second, recruitment was from a single specialist university centre, which limits generalisability; participants were also followed within a publicly funded programme in which cost is not a barrier, so the findings do not transfer directly to systems with patient cost-sharing. Third, although the response rate was 77%, no information was available on the patients who declined to take part, so the extent of any non-response bias cannot be assessed. Fourth, participation required internet access and adequate digital literacy, which is likely to have selected a comparatively younger and more digitally engaged subgroup, and the requirement of being under active programme follow-up is likely to have selected a comparatively more adherent one. Fifth, adherence was self-reported rather than measured objectively, and self-report is subject to recall and social-desirability bias, both of which act in the direction of overestimating adherence. Sixth, a pronounced ceiling effect was present in both the ACDS (22.4% at the maximum) and the CAT (34.5% at the maximum), restricting variance and attenuating the associations that can be detected. Seventh, the internal consistency of the ACDS in this sample (α = 0.562) was substantially lower than in its Polish validation study and below the value conventionally regarded as acceptable for group-level comparisons,⁴³ which weakens confidence in the total score as a unidimensional measure of adherence. Eighth, information on individual dosing schedules, recent treatment switching and treatment interruptions was not collected, and the injectable category comprises agents which markedly differ in administration frequencies. Ninth, several established correlates of adherence were not measured with validated instruments, including cognitive functioning, depression, health literacy, disease activity and the extent of shared decision-making; the accompanying-symptom checklist reported in Table 2 is a self-reported screen and is not a substitute for formal assessment, so residual confounding cannot be excluded. Tenth, although the EDSS was scored by the treating neurologist, it was entered into the questionnaire by the patient and was not subsequently checked against the medical record, so transcription error cannot be excluded. Finally, the CAT and the medical care domain of the SATMED-Q partly overlap conceptually; although this did not produce multicollinearity, it may still contribute to the observed pattern of associations.

The results of our study suggest that efforts to improve adherence in MS may need to extend beyond the pharmacological properties of DMTs to encompass the quality of the therapeutic relationship. Strengthening physician-patient communication, promoting shared decision-making and systematic assessment of treatment satisfaction are clinically reasonable objectives in their own right, and both were positively associated with self-reported adherence in the present data, although no beneficial effect on adherence was demonstrated. Because this study did not test any such interventions, these suggestions should be regarded as hypotheses for future longitudinal and interventional research rather than as evidence-based recommendations. Studies that combine self-reported measures with objective adherence data, complemented with formal assessment of cognition and mood, as well as disability scores taken directly from the clinical records, would be particularly informative.

Conclusions

In this single-centre cross-sectional sample of patients with relapsing-remitting multiple sclerosis, self-reported adherence to disease-modifying therapies was intermediate to high, with fewer than half of the participants meeting the criterion for high adherence. Higher self-reported treatment satisfaction and more favourable ratings of physician-patient communication were associated with modestly higher adherence scores, and greater disability was associated with lower adherence scores. Because of the cross-sectional design and the exclusive use of self-reported measures, only associations between adherence and other analyzed factors could be revealed, with no conclusions drawn about directionality or causality links. Longitudinal and interventional studies are required to establish the kind of intervention (including communication-focused or shared decision-making ones) which can improve adherence to disease-modifying therapies in MS.

Funding Statement

This work was supported by Wroclaw Medical University (SUBZ.A520.26.030).

Data Sharing Statement

The de-identified verified dataset and the analysis scripts supporting the findings of this study are available from the corresponding author on reasonable request.

Ethics Approval and Consent to Participate

The ethical opinion was provided by the Bioethics Committee of Wroclaw Medical University (approval no. KB 175/2022). It was performed in accordance with the Helsinki Declaration. Each patient gave informed electronic consent.

Disclosure

The authors declare that they have no financial or non-financial competing interests in relation to this work.