Effects of horizontal versus vertical switching of disease-modifying treatment after platform drugs on disease activity in patients with relapsing–remitting multiple sclerosis in Austria

Objectives To compare in a nationwide observational cohort the effectiveness, frequency and reasons for treatment interruption of dimethylfumarate (DMF) and teriflunomide (TERI) (horizontal switchers) versus alemtuzumab (AZM), cladribine (CLAD), fingolimod (FTY), natalizumab (NTZ), ocrelizumab (OCR) and ozanimod (OZA) (vertical switchers) in patients with relapsing–remitting multiple sclerosis (pwRRMS) and prior interferon beta (IFN-beta) or glatiramer-acetate (GLAT) treatment. Materials and methods The “horizontal switch cohort” included 669 and the “vertical switch cohort” 800 RRMS patients. We used propensity scores for inverse probability weighting in generalized linear (GLM) and Cox proportional hazards models to correct for bias in this non-randomized registry study. Results Estimated mean annualized relapse rates (ARR) were 0.39 for horizontal and 0.17 for vertical switchers. The incidence rate ratio (IRR) in the GLM model showed an increased relapse probability of 86% for horizontal versus vertical switchers (IRR = 1.86; 95% CI 1.38–2.50; p < 0.001). Analyzing the time to the first relapse after treatment switch by Cox regression, a hazard ratio of 1.58 (95% CI 1.24–2.02; p < 0.001) indicated an increased risk of 58% for horizontal switchers. The hazard ratios for treatment interruption comparing horizontal versus vertical switchers were 1.78 (95% CI 1.46–2.18; p < 0.001). Conclusions Horizontal switching after a platform therapy resulted in a higher relapse and interrupt probability and was associated with a trend towards less EDSS improvement comparing to vertical switching in Austrian RRMS patients.

We here, thus, sought to compare the effectiveness, frequency and reasons for treatment interruption in patients with prior IFN-ß or GLAT treatment switching to either DMF and TERI (horizontal switchers) or AZM, CLAD, FTY, NTZ, OCR and OZA (vertical switchers) in a nationwide observational cohort using prospectively collected data from a real-life setting.

Data collection
The Austrian MS Treatment Registry (AMSTR) [17,18], established in 2006 to maintain quality control and comply with reimbursement regulations of the Austrian sick funds, allows to obtain clinical data, assess indications, clinical profiles of treated patients and to monitor safety in real life. The AMSTR is part of the dense network of MS centers in Austria, which is constituted by MS clinics from neurological departments and some dedicated neurological practices that have been assigned this status by the Austrian Society of Neurology based on defined quality criteria. In addition, prescriptions of disease-modifying therapies (DMTs) for MS are exclusively reserved for MS centers. Thus, prescriptions and treatment documentations are evenly distributed across Austria. The AMSTR is compliant with Austrian laws on bioethics and it was also approved by the ethical committee of the Medical University of Vienna (EC number 2096/2013).
The AMSTR documents anonymous baseline data, including the date of clinical onset of MS and disease duration, relapses in the prior 12 months, EDSS, magnetic resonance imaging (MRI) activity and previous DMT. Follow-up data (relapses, EDSS, adverse events [AEs], change or discontinuation of treatment) are required to be documented every 3-6 months. Because of the structure and requirements of the AMSTR, MRI results were only available at baseline before start of treatment, but unfortunately not during follow-up. Each relapse has to be confirmed by a neurologist at the MS center and documented in the AMSTR. Documentation also requires date of relapse onset, EDSS and use/dose of i.v. methylprednisolone treatment. Besides the fact that applying the AMSTR is mandatory for reimbursement, external and independent data monitoring to improve data acquisition, input and management in terms of completeness and plausibility of documented data constitutes a special quality-related feature of the AMSTR.
We identified two cohorts of patients with relapsing-remitting multiple sclerosis (RRMS) from the AMSTR since 2006, who have either switched from IFNbeta and/or GLAT to DMF and TERI ('horizontal switchers') or to AZM, CLAD, FTY, NTZ, OCR and OZA ('vertical switchers'). The horizontal switchers correspond to lateral switchers and the vertical switchers to escalation treatment as cited in the literature [33]. Patients had to remain on the switched treatment for at least three months due to disease activity, patients preference and side effects. The horizontal switch cohort included 669 and the vertical switch cohort 800 RRMS patients.

Outcome measures
The primary outcome measure was the annualized relapse rate (ARR) under treatment during this period. Relapses were defined as new or worsening neurological symptoms lasting for at least 24 h in the absence of fever. Further outcome measures were time to first relapse and sustained disability progression or regression. Sustained disability progression or regression was defined as an increase or decrease from baseline of at least 1.0 point in the EDSS (or at least 0.5 points for patients with a baseline EDSS score greater than 5.5) that persisted for at least 12 or 24 weeks.
For analyses of treatment interruption, we defined two causes, namely (a) stopping treatment as permanent treatment interruption in the AMSTR and (b) switching treatment as treatment interruption and starting with a new medication in the AMSTR.

Statistical methods
All effects estimated in comparing treatment groups were average treatment effects (ATE). To control the bias for nonrandomized assignment to the treatment groups, we used normalized inverse probability weighting (IPW). Propensity scores for treatment in the horizontal and vertical switchers were estimated with a logistic regression model with the baseline parameters age, duration of disease, number of relapses 12 months prior to baseline, EDSS, presence of at least 9 cerebral MRI T2 lesions and at least one contrast-enhancing brain MRI lesion, and previous therapy as independent variables. These variables were included in the model because of their clinical meaning, independent from their significance as a predictor in the model. Therefore, we tried to overcome the problems of being misled by false-positive predictors in a multiple testing situation as well as of missing relevant variables by abandoning them in a beta failure decision. The study cohort was truncated for two patients producing weights higher than ten. Treatment groups were well balanced for all variables after weighting. The vertical switcher group produced higher propensity scores for their treatment on average, leading to a right tail of the propensity score distribution closer to one and a left tail shorter against zero compared to the propensity score distribution of the horizontal switcher group. This fact produced higher weights on the left and right sides of the propensity score distribution for the horizontal switcher group, leading to a higher sample size for the horizontal group after weighting. A generalized linear model (GLM) with relapse count as Poisson distributed-dependent variable and log-transformed observation time in years as offset variable was used to estimate the treatment effect on the ARR in the observation period.
Augmented inverse probability weighting was used to analyze the change of EDSS from baseline to the last visit in the observation period, with the difference between last visit and baseline (negative as improvement, positive as worsening) for each patient as dependent variable. Potential means for these changes were estimated for each treatment from the model.
We used Cox proportional hazards models for analyzing EDSS progression and regression confirmed after 12 and 24 weeks, and the relapse hazard in the observation period.
All models included treatment as categorical factor and normalized inverse propensity scores as weights regarding the survey character of the study. All variables used for propensity scoring were also used in the outcome models as independent variables to obtain adjusted treatment effects. We applied this double robust approach, because the ATE estimator remains consistent if at least one of the two, the propensity score model or the outcome regression, is specified properly. Thus, the misspecification of only one of the two models would not cause any harm to the ATE estimator [35].
The proportional hazards assumption for the Cox models had been verified by non-significant deviations from the proportional hazards assumption for each covariate in the model using Schoenfeld residuals.

Results
According to the pre-defined inclusion criteria, the cohort of horizontal switchers included 669 (DMF: 223, TERI: 446 patients) and the cohort of vertical switchers 800 (AZM: 16, CLAD: 27, FTY: 523, NAT: 205, OCR: 21, OZA: 8 patients) RRMS patients, respectively (Table 1). The baseline data of both groups before treatment switch are summarized in Table 2 and show certain imbalances for some baseline variables. Vertical switchers were younger, had a shorter duration of MS and higher values concerning the EDSS, relapse rate and T2-hyperintense and Gd-enhancing T1 lesion loads. Normalized IPW resulted in a weighted sample size of 740 and 698 patients, respectively, and this was well balanced indicated by standardized mean differences (SMD) below 0.1 ( Table 2). Mean follow-up was 2.8 years (SD 2.0) for unbalanced horizontal and 3.6 years (SD 2.3) for unbalanced vertical switchers, after IPW 2.7 (SD 1.9) and 3.6 (SD 2.2), respectively.
Mean annualized relapse rates of the unbalanced cohorts (ARR) were 0.23 (SD 0.61) for horizontal and 0.19 (SD 0.46) for vertical switchers and after IPW 0.39 (SD 0.9) and 0.17 (SD 0.42), respectively. The incidence rate ratio (IRR) in the GLM model for relapses showed an increased relapse  (Fig. 1, Table 3).
The reasons for switching medication within the horizontal group were mainly disease progression (n = 93), patients' wishes (n = 66), AEs (n = 44), and within the vertical group the reasons were disease progression (n = 82), patients' wishes (n = 55) and AEs (n = 30). It is important to note that treating neurologists were allowed to name several reasons per patient.
Pregnancy or the wishes to conceive were documented in 11 patients in the horizontal and 25 patients in the vertical cohort.

Discussion
In this observational study across a nation with welldefined MS treatment units, we analyzed prospectively collected real-world data to compare the effectiveness of change in treatment, frequency and reasons for treatment interruption of DMF and TERI (horizontal switchers) versus AZM, CLAD, FTY, NTZ, OCR and OZA (vertical switchers) after IFN-beta and GLAT therapy in large cohorts of RRMS patients. By nature, the different approved treatment labels caused differences in baseline characteristics of the cohorts (Table 2). To control for these, we used normalized IPW and provided well balanced baseline parameters ( Table 2).
There is some discrepancy concerning studies comparing a horizontal and vertical DMT switch [33,34]. D'Amico et al. analyzed retrospectively 48 RRMS patients with a horizontal switch (IFN-beta and GLAT) and 43 patients with an escalation therapy (FTY, NTZ, cyclophosphamide) after experiencing on-treatment disease activity. No significant differences regarding NEDA 3, in timing to reach the first relapse and obtaining an EDSS of 4.0 were found [33]. In contrast, Chalmer et al. retrieved data from The Danish Multiple Sclerosis Registry on RRMS patients who experienced disease breakthrough on moderately effective therapy (IFN-beta, GLAT, TERI) and switched either to another moderately effective treatment (IFN-beta, GLAT, TERI and DMF) or to a highly effective therapy (FTY and NTZ). Both treatment groups were balanced by propensity score matching and comprised 404 patients. ARR (0.22; IQR 0.19-0.27 versus 0.32; IQR 0.28-0.37) and hazard of reaching a first relapse (HR 0.65; 95% CI 0.53-0.80) were significant lower after a vertical switch. Additionally, no evidence of disability worsening and a weak evidence of disability improvement were found [34].
Comparing our results, we also identified a higher relapse probability and a trend towards less EDSS improvement with a horizontal switch after a platform therapy comparing to a vertical switch. Differences to the aforementioned studies also comprised a treatment switch because of AEs and patients' wishes. Further, we also included AZM, CLAD, OCR and OZA as a vertical swich therapy. Importantly, in comparison with the aforementioned studies, we analyzed the by far largest study population with a total of 1469 RRMS patients.
Several studies so far demonstrated that patients may stay stable after switching (for various reasons) from IFNbeta and GLAT to DMF and TERI [23][24][25][26][27][28]. In particular, Buron et al. provided data from the Danish Multiple Sclerosis Registry showing that switching from injectables to DMF and TERI in clinically stable RRMS patients did not increase the risk of disability accumulation [27]. On the contrary, various studies revealed a superior efficacy of highly effective therapies in scenarios where a vertical switch had been performed due to suboptimal response to prior DMTs [29][30][31]. This contrasts the findings of D'Amico et al. who observed similar efficacy of a lateral switch to IFN-beta compared to an escalation switch to FTY in RRMS patients who failed with other injectables [32].
The hazard ratio for treatment interruption comparing a horizontal versus vertical switch was significantly higher for horizontal switchers. The number of patients interrupting treatment was 263 (39.3%) in the horizontal cohort (152 stopped, and 111 switched) and 296 (37%) within the vertical group (165 stopped and 131 switched). The reasons for permanent treatment interruption within the horizontal group were mainly patients' wishes and AEs, while within the vertical cohort, respective reasons were patients' wishes and disease progression. In contrast, the reasons for switching medication within the horizontal and vertical groups were mainly disease progression. These results are in line with various studies evaluating reasons for a switch and treatment discontiuation [18,[21][22][23][24]. He et al. also showed a lower hazard of treatment discontinuation (HR 0.55, 95% CI 0.31-0.98) with a vertical (FTY) versus a horizontal (IFNbeta) switch [31]. In our study, almost 80% of patients in the horizontal cohort switched to highly effective therapy, whereas 20% deescalated to a moderately effective treatment within the vertical cohort. In contrast, Patti et al. and Papp et al. documented a lower escalating rate of 53-60% after treatment with DMF and TERI [21,23]. Furthermore, in 27.3-39.1% of Italian RRMS patients, treatment was deescalated to IFN-beta, GLAT, TERI and DMF after NTZ and FTY therapy in the SWITCH study [21].
The ideal timepoint during a patient's disease course to start treatment with a DMT or to switch a DMT is difficult to predict. Ziemssen et al. proposed two windows of opportunity for treatment optimization in MS [36]. He postulated that both early initiation of treatment and prompt intervention if disease activity persists despite initial treatment are critical to optimization of treatment outcomes [36]. In line with these notions, several studies provided evidence of superior efficacy of early intensive treatment compared to escalation treatment [37][38][39].
The major strength of our study is represented by the fact that these analyses are based on a nationwide observational data collection from approved MS facilities, comprising patients in Austria who have been treated with DMF and TERI or AZM, CLAD, FTY, NTZ, OCR and OZA after a treatment with IFN-beta and/or GLAT. The AMSTR represents a secure web-based platform that enables treating neurologists in all Austrian MS centers to immediately perform online documentation during patient visits. To ensure high documentation and data quality in terms of completeness and plausibility, the AMSTR is monitored by an external and independent clinical research organization. Our real-world data show a low ARR, progression rate and discontinuation rate for all drugs reflecting high quality maintenance of MS patients in Austria (data not shown).
On the other hand, an important limitation of our study has to be considered as well when interpreting our results. Information on MRI findings were only available at baseline before start of treatment and we included these as an independent variable for propensity scoring and in the respective outcome models. Information on respective findings during follow-up unfortunately is missing. Furthermore, the reasons for horizontal versus vertical switching could be systematically different in a way not captured by the propensity score, and these cannot be captured by a data-base as used in the AMSTR.
In conclusion, we here present the first large analysis comparing horizontal (DMF and TERI) versus vertical switchers (AZM, CLAD, FTY, NTZ, OCR and OZA) of pwRRMS after a treatment with IFN-beta and/or GLAT. Reasons for switching were manifold including suboptimal treatment response, AEs and convenience. A horizontal switch after a platform therapy resulted in a higher relapse probability and discontinuation rate and a trend towards less EDSS improvement compared to a vertical switch. Therefore, switching to a higher efficacious treatment should be the preferred treatment choice in the case of pwRRMS who do not adequately respond to platform therapies. Given the currently available armamentarium of treatments and considering the findings of our study, a horizontal switch among platform therapies currently shall only be considered in case of side effects of the previous platform therapy.