Delay From Treatment Start to Full Effect of Immunotherapies for Multiple Sclerosis

Izanne Roos; Emmanuelle Leray; Federico Frascoli; Romain Casey; J. William L. Brown; Dana Horakova; Eva K. Havrdova; Maria Trojano; Francesco Patti; Guillermo Izquierdo; Sara Eichau; Marco Onofrj; Alessandra Lugaresi; Alexandre Prat; Marc Girard; Pierre Grammond; Patrizia Sola; Diana Ferraro; Serkan Ozakbas; Roberto Bergamaschi; Maria José Sá; Elisabetta Cartechini; Cavit Boz; Franco Granella; Raymond Hupperts; Murat Terzi; Jeannette Lechner-Scott; Daniele Spitaleri; Vincent Van Pesch; Aysun Soysal; Javier Olascoaga; Julie Prevost; Eduardo Aguera-Morales; Mark Slee; Tunde Csepany; Recai Turkoglu; Youssef Sidhom; Riadh Gouider; Bart Van Wijmeersch; Pamela McCombe; Richard Macdonell; Alasdair Coles; Charles B. Malpas; Helmut Butzkueven; Sandra Vukusic; Tomas Kalincik

Disclosures

Brain. 2020;143(9):2742-2756. 

In This Article

Materials and Methods

Ethics statement

The MSBase registry (Butzkueven et al., 2006) (registered with WHO ICTRP, ID ACTRN12605000455662) was approved by the Melbourne Health Human Research Ethics Committee, and by the local ethics committees in all participating centres (or exemptions granted, according to applicable local laws and regulations). Written informed consent was obtained from enrolled patients as required. The OFSEP cohort (Vukusic et al., 2020) (registered with WHO ICTRP, ID NCT02889965) was collected with approval from and in accordance with French Commission Nationale Informatique et Libertés and French law relative to observational research.

Patients

Longitudinal clinical and demographic data from 125 centres in 37 countries were extracted from the MSBase registry in November 2018 and from 39 French centres in the OFSEP cohort in December 2018. One additional non-MSBase non-OFSEP centre from Cambridge was included in the MSBase cohort (only patients given alemtuzumab) (Tuohy et al., 2015). The following inclusion criteria were applied prior to enrolment: diagnosis of multiple sclerosis or clinically isolated syndrome as per the 2005 or 2010 McDonald criteria (Polman et al., 2005, 2011), commencement of and persistence on a disease modifying therapy (DMT) for at least 12 months, minimum 3-year pretreatment follow-up, yearly follow-up during the treatment epoch (defined below) and presence of the minimum dataset. Patients were diagnosed with clinically isolated syndrome at the time of start of their treatment, and with ≥3 year follow-up from their first symptom. The minimum dataset consisted of patient age, sex (categorized as female and male), multiple sclerosis duration at baseline, disease phenotype (clinically isolated syndrome, relapsing-remitting, secondary progressive, primary progressive, progressive-relapsing), disability information [quantified with Expanded Disability Status Score (EDSS)] (Kurtzke, 1983) at baseline and two subsequent time points at least 6 months apart and, where applicable, date of treatment cessation.

Baseline was defined as the start of the index DMT (including a new therapy or treatment switch). The prospective follow-up period was defined as the time from the first to the last available EDSS. A treatment epoch was defined as time including 3 years prior to baseline and 1 year (for the effect on relapses) and 5 years (for the effect on disability; see below) after baseline.

All available and sufficiently represented DMTs were included in the analysis. A therapy was viewed as sufficiently represented for application of the method described below in 'Proof of principle: the method to identify therapeutic lag' section if more than 200 events (relapses or progression-of-disability events) occurred during the analysed treatment epoch. Duration of treatment effect after the last dose was estimated according to pharmacodynamics, clinical experience or previous evidence (Stellmann et al., 2017), as follows [in keeping with our previous work (Kunchok et al., 2020)]: 4 years after the last dose for alemtuzumab (Coles et al., 2017), 14 days for dimethyl fumarate, 7 days for all interferon therapies and glatiramer acetate (Stellmann et al., 2017), 30 days for fingolimod (David et al., 2012), 180 days for mitoxantrone (Hartung et al., 2002), 60 days for natalizumab (Sheremata et al., 1999), 270 days for rituximab or ocrelizumab, and 5 years for autologous haematopoietic stem cell transplant (Sormani et al., 2017). Treatment with DMTs was only allowed as a monotherapy. The two treatment epochs were merged if a period of DMT interruption, with subsequent commencement of the same therapy, was shorter than the duration of treatment effects. Treatment gaps exceeding the abovementioned periods were recorded as separate treatment epochs. In patients in whom multiple eligible baselines were identified multiple treatment epochs per patient were studied. Multiple epochs per patient were treated as independent.

All data were prospectively collected during routine clinical care predominantly from tertiary multiple sclerosis centres (Kalincik and Butzkueven, 2019; Vukusic et al., 2020). Information was entered near real-time (usually at the time of a clinic visit) into the iMed patient record or online MSBase data entry system for MSBase or EDMUS patient record for OFSEP (Confavreux et al., 1992). Data were subject to standardized data quality processes (Supplementary Table 1) (Kalincik et al., 2017; Vukusic et al., 2020).

Study Outcomes

This study evaluated the time from treatment start to its full clinically manifest effect on relapses and disability progression events.

Relapses were defined as the occurrence of new symptoms or the exacerbation of existing symptoms for at least 24 h in the absence of concurrent illness or fever, and occurring at least 30 days after a previous relapse (Schumacher et al., 1965). In the primary analysis, relapses were analysed as recorded by the treating neurologist.

Confirmed progression of disability was defined as an increase in EDSS by 1.5 steps if baseline EDSS was 0, increase by 1 step if baseline EDSS was between 1 and 5.5 or increase in EDSS by 0.5 step if baseline EDSS was above 5.5, confirmed at least 6 months later (in the absence of a relapse within 30 days prior to the confirmatory EDSS) and sustained for the remainder of the treatment epoch (Kalincik et al., 2015). The pretreatment baseline EDSS was as documented at the first recorded visit and re-baselined at the commencement of the index DMT. Progression of disability independent of relapse activity was defined as 6-month confirmed progression of disability (see above), where the increase in disability could not be attributed to a preceding relapse (Lorscheider et al., 2019). This was ensured by the absence of a recorded relapse between the EDSS leading to the progression-of-disability event and the most recent preceding EDSS.

The minimum on treatment follow-up period differed according to the studied outcome. For the analysis of the time to the full effect of DMTs on relapses, patients required a minimum of 1 year on-treatment follow-up. All relapses recorded during this year were used in the analysis. For analysis of disability outcomes, patients were treated for at least 1 year and treatment epochs of 5 years post-baseline were analysed on an 'intention-to-treat' basis. This means that all disability progression events recorded during the 5-year period were analysed, irrespective of the actual treatment status. These differences in analytical approaches are motivated by the observation that the effect of DMTs on relapses are short term, whereas the effect on disability is cumulative (Brown et al., 2019). This resulted in two distinct, but overlapping, cohorts for the analysis of the two studied outcomes.

Statistical Analysis

All analyses were performed using R version 3.5.3. Point and interval estimates of distribution were expressed as means with 95% confidence intervals (CIs), or medians with quartiles, as appropriate. All hypotheses were tested with a two-tailed 0.05 level of significance.

Proof of Principle: The Method to Identify Therapeutic Lag

The MSBase cohort was used as a discovery cohort to develop the method to identify the duration of therapeutic lag in the effect of DMTs on relapse events. Patients diagnosed with remitting relapsing multiple sclerosis or clinically isolated syndrome (i.e. patients most likely to experience relapses) were selected for this analysis. For each DMT a density curve of the relapse events during the treatment epoch (3 years before and 1 year after baseline) was produced. The Sheather-Jones criterion was applied for optimal estimation of kernel density and bandwidth selection of the density curve (Sheather and Jones, 1991).

The density curves were then used to identify the first local minimum (the minimum incidence of relapses) after commencing an index DMT, by calculating the first derivatives of the curves (Figure 1A). This local minimum translates to the time point at which stabilization of the effect treatment on the measured outcome is reached, therapeutic lag for relapses (Tr). A 200-event minimum for study inclusion was guided by analyses showing that Tr is not identifiable for any DMT below this threshold. The estimates of Tr were recalculated by (i) non-parametric bootstrap with 10 000 repetitions; and (ii) Monte Carlo simulations using 80% of the cohort, without replacement, and 10 000 repetitions in order to model their probability distributions. No substantial differences between the two estimates of Tr were found. As the Monte Carlo method is more conservative and makes no assumption that the sample is an estimate of the true population, we chose to use the Monte Carlo method for the remainder of analyses (principles of Monte Carlo simulations are described in Supplementary Figure 1). Gaussian mixture models were used to identify the point associated with the maximum density probability where the simulations resulted in multimodal distributions of the sought points (McLachlan and Peel, 2000). This estimate of the point was confirmed by calculation of the point from the entire available population.

Figure 1.

Density curve and first derivative of relapse (A) and 6-month confirmed disability progression (B) events prior to and after the commencement of natalizumab. Time point 0 indicates the start of index therapy. The first post-treatment local minimum of the first derivative, representing the point of stabilization of treatment effect, is indicated in purple. The duration of therapeutic lag on relapses (Tr) and disability progression (Td) is indicated by double headed horizontal arrows. The last stable point prior to start of the index therapy, identified by the last pretreatment local minimum of the first derivative, is shown in orange. Pretreatment and on-treatment ARRs were calculated in the periods indicated by dashed arrows.

The OFSEP cohort was used to perform an external validation analysis of the proposed method to detect Tr. The validation used the same inclusion criteria and methodology as the discovery analysis (see above). The differences in mean Tr between the MSBase and the OFSEP cohorts for each DMT were estimated with bootstrap analyses, including their bias corrected and accelerated 95% CIs.

To explore effects of the shape of the density curve on Tr, stability analyses were conducted. Here, the combined MSBase-OFSEP cohort was used. First, the association between the height of the peak in relapse density immediately prior to baseline (driven by relapses preceding commencement of index DMTs) and Tr was studied. Furthermore, in a series of simulations, Tr was evaluated in random samples of patients from the combined cohort, with sequentially increasing number of total recorded relapses in each sample.

Therapeutic Lag for Relapses. To estimate therapeutic lag for relapses Tr, we have used the combined MSBase-OFSEP cohort and the method established above. Monte Carlo simulations were used to estimate the 95% CIs of Tr. An additional point was identified as the point of the last 'stable' density of relapses prior to the peak of relapse incidence that prompted initiation of index DMTs (last local minimum of the first derivative before treatment start) (Figure 1A). This point was used to calculate the pretreatment annualized relapse rates (ARRs) as the relapse density divided by the cumulative follow-up time prior to that point. The on-treatment ARR was calculated from Tr to the end of treatment epoch, dividing relapse density by the respective cumulative follow-up time. The pre- and on-treatment ARRs and their 95% bias corrected and 95% confidence interval were then visualized.

Sensitivity analyses were performed with a more stringent definition of relapses, only including events treated with corticosteroids or resulting in a change in EDSS.

Therapeutic Lag for Disability Progression. The method described in 'Proof of principle: the method to identify therapeutic lag' was used to analyse therapeutic lag for disability progression (Td) during the 5-year post-baseline period in patients who were treated for the minimum of 1 year from the merged MSBase-OFSEP cohort (Figure 1B). Patients with both relapsing and progressive multiple sclerosis forms were included in this analysis. An additional analysis studied therapeutic lag for disability progression independent from relapses (TPIRA).

Finally, sensitivity analyses evaluated the robustness of results for differential treatment persistence and follow-up (2, 3 and 5 years).

Data Availability

MSBase is a data processor, and warehouses data from individual principal investigators who agree to share their datasets on a project-by-project basis. Data access to external parties can be granted upon reasonable request at the sole discretion of each OFSEP and MSBase Principal Investigator (the data controllers), who will need to be approached individually for permission.

processing....