A web-based decision support tool for prognosis simulation in multiple sclerosis

Multiple Sclerosis and Related Disorders (]]]]) ], ]]]–]]]

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/msard

A web-based decision support tool for prognosis simulation in multiple sclerosis Mário Veloson ARN – Anestesia, Reanimação e Neurologia – Lda, Campo Grande 14 – 6ºA, 1700-092 Lisboa, Portugal Received 16 December 2013; received in revised form 14 April 2014; accepted 22 April 2014

1.

KEYWORDS

Abstract

Multiple sclerosis; Decision support; Simulation model; Prognosis; Web-based tool; Disease progression assessment

A multiplicity of natural history studies of multiple sclerosis provides valuable knowledge of the disease progression but individualized prognosis remains elusive. A few decision support tools that assist the clinician in such task have emerged but have not received proper attention from clinicians and patients. The objective of the current work is to implement a web-based tool, conveying decision relevant prognostic scientific evidence, which will help clinicians discuss prognosis with individual patients. Data were extracted from a set of reference studies, especially those dealing with the natural history of multiple sclerosis. The web-based decision support tool for individualized prognosis simulation was implemented with NetLogo, a program environment suited for the development of complex adaptive systems. Its prototype has been launched online; it enables clinicians to predict both the likelihood of CIS to CDMS conversion, and the long-term prognosis of disability level and SPMS conversion, as well as assess and monitor the effects of treatment. More robust decision support tools, which convey scientific evidence and satisfy the needs of clinical practice by helping clinicians discuss prognosis expectations with individual patients, are required. The web-based simulation model herein introduced proposes to be a step forward toward this purpose. & 2014 Elsevier B.V. All rights reserved.

Introduction

history of the disease is essential to realizing the potential long-term implications of any intervention, including treatment (Vukusic and Confavreux, 2007). However, the individualized prognosis of multiple sclerosis remains uncertain and continues to be a major challenge in the clinical practice of neurologists. In recent years, several tools have been developed to assist clinicians and patients formulate custom prognostic expectations (Kister et al., 2013; Galea et al., 2013; Trojano et al., 2002; Bergamaschi et al., 2007; Hughes et al., 2012). Despite the importance of an individualized prognosis, these

Although the clinical manifestations, the evolution, and the pathology of multiple sclerosis are heterogeneous, studies of the natural history give valuable insight into the temporal course of relapses, disease progression, and predictive clinical factors (Tremlett et al. 2010). Thus, the natural n

Mobile: +351 964020950; fax: +351 217993913. E-mail address: [email protected]

http://dx.doi.org/10.1016/j.msard.2014.04.005 2211-0348/& 2014 Elsevier B.V. All rights reserved.

Please cite this article as: Veloso M. A web-based decision support tool for prognosis simulation in multiple sclerosis. Multiple Sclerosis and Related Disorders (2014), http://dx.doi.org/10.1016/j.msard.2014.04.005

2

M. Veloso

models, as many other prognostic models developed in various medical fields, have rarely been used in clinical practice (Wyatt and Altman, 1995; Dong et al., 2012). Among the reasons for low adoption, the dissociation of these models from the needs of clinicians and their failure to provide information that is relevant to the decision process in a comprehensive and easy to use format should be emphasized (Heesen et al., 2011). The main objective of this study is to implement a tool that is easily accessible, simple to use, and satisfies neurologists' needs in clinical practice, such that it becomes actually used. To achieve this objective, a web-based computer prognostic simulation model was developed to allow neurologists to discuss with the individual patients:

 the likelihood that the patient with a clinical isolated  

syndrome (CIS) should convert to clinically definite multiple sclerosis (CDMS); the long-term prognosis (over the severity of the disease and the transition probability for the secondary progressive phase (SPMS)); and the need to treat and/or modify therapy.

visualization of simulations of agent based/complex adaptive systems. The current model adapts, reformulates, and extends a previously described simulation model (Veloso, 2013), and is incorporated into a web page as a Java applet. A different modelling approach, using distinct algorithms, was adopted for the implementation of the current model, enabling the utilization of a variety of reference studies addressing common issues. This strategy contributed to enlarge the model's scope, now also including patients with CIS, to improve the user interface, and to improve the model validation by comparing evidence data from different studies. The data used to implement the model were extracted from a set of reference studies, especially those dealing with the natural history of multiple sclerosis. Given the multitude of clinical and para-clinical factors described in the literature as predictive of disease progression, only a subset of relevant factors was considered for this model, so that the concepts underlying this research could be validated. Selection criteria of the prognostic factors used by the model included:

 Utilization of data variables that are used in the clinical 2.

Methods



The simulation model was developed with NetLogo 5.0.4 (Wilensky, 1999), which provides a particularly powerful programming environment for the implementation and Table 1 Objectives



practice and that have been published in relevant scientific journals. Published results in a format following the design of the model. The minimum set of factors identified as the best answer to the questions posed in the objectives of the study.

Prognostic factors used by the model. Used prognostic factors/data source articles

CIS-CDMS conversion 10 Years evolution Baseline MRI (normal vs abnormal)/(O’Riordan et al., 1998) 10 Years evolution CSF (OCB + vs OCB )/(Dobson et al., 2013) 20 Years evolution Baseline MRI (normal vs abnormal)/(Fisniku et al., 2008) RRMS-SPMS conversion 10 Years evolution Baseline MRI (normal vs abnormal)/(O’Riordan et al., 1998) 10 Years evolution CSF (OCB + vs OCB )/(Amato and Ponziani, 2000) 20 Years evolution Baseline MRI (Normal vs Abnormal)/(Fisniku et al., 2008) 10 & 20 Years Age at disease onset, no relapses first 2 years/(Scalfari et al., 2013) evolution 10 & 20 years BREMS score: age at onset of disease, gender, sphincter onset, pure motor onset, motor-sensory onset, conversion sequel after onset, number of involved functional systems at onset, sphincter plus motor relapses, EDSS Z4/(Bergamaschi et al., 2007; Bergamaschi et al., 2012) Levels of disability 10 Years evolution Disability curves: EDSS & nº of years/(Achiron et al., 2003) 20 Years evolution Baseline MRI (Number of lesions) (Fisniku et al., 2008) 20 Years evolution MSSS: EDSS & nº of years/(Roxburgh et al., 2005) Therapeutic impact CIS: 2 years BENEFIT study/(Kappos et al., 2006) evolution RRMS:20 years BREMS score/(Bergamaschi et al., 2012) evolution Treatment optimization selected year MRI (new lesions, relapses (number), EDSS)/(Zaffaroni, 2005) EDSS: Expanded Disability Status Scale; MSSS: Multiple Sclerosis Severity Score; BREMS: Bayesian Risk Estimate for Multiple Sclerosis; CIS: Clinical Isolated Syndrome; RRMS: Relapsing-Remiting Multiple Sclerosis; CDMS: Clinical Definite Multiple Sclerosis; SPMS: Secondary Progressive Multiple Sclerosis; OCB: Oligoclonal Bands; MRI: Magnetic Resonance Imaging; CSF: Cerebrospinal fluid. Please cite this article as: Veloso M. A web-based decision support tool for prognosis simulation in multiple sclerosis. Multiple Sclerosis and Related Disorders (2014), http://dx.doi.org/10.1016/j.msard.2014.04.005

Web-based decision support tool in MS

3

Selected variables, their scope, and the source articles from which the data were obtained, are outlined in Table 1. Although the results presented by the model simulations reproduce validated data, as published in the source articles herein reviewed, additional validation has been performed using a dataset of 50 patients with relapsingremitting multiple sclerosis (RRMS) and at least 10 years of disease evolution. The characteristics of this patient Table 2

Validation group of 50 patients. Validation group (N =50)

Gender Male – N (%) 13 (26%) Female – N (%) 37 (74%) Age at disease onset Mean (SD) 28.42 (78.39) Median 25.65 Functional system affected – N (%) Motor 12 (24.0%) Sensory 18 (36.0%) Coordenation 1 (2.0%) Visual 14 (28.0%) Brainstem 5 (10.0%) Undefined 0 Interval 1st–2nd relapse (no of years) Mean (SD) 4.66 (7 6.03) Median 2.59 Years of follow up Mean (SD) 17.01 (7.86) Median 14,52 EDSS At 10 years (N =50) Mean (SD) 2.02 (1.53) Median 2 MSSS Mean (SD) 2.4 (2.03) Median 2.34

3.

At 20 years (N =17) 3.41 (2.56) 2.5 2.71 (2.69) 1.29

N: Number of patients; SD: standard deviation; EDSS: Expanded Disability Status Scale; MSSS: Multiple Sclerosis Severity Score; BREMS: Bayesian Risk Estimate for Multiple Sclerosis.

Table 3

dataset have been previously described (Veloso, 2013) and are outlined in Table 2. Moreover, a few components of the current model have already been validated in the previously referred work. The additional validation focused on the model's ability to accurately classify the disability level of the patients ((i) low disability corresponding to EDSS o= 3; (ii) moderate disability corresponding to EDSS score greater than three and less than six; (iii) severe disability when EDSS4 = 6) at 10 and 20 years of evolution. MS curves (Hughes et al., 2012), a multiple sclerosis severity rank calculator online tool, available at the MSBase platform, was used for comparison purposes of the model's accuracy.

Results

The key aspect of the model is its ability to simulate interactively the long term progression of the disease for any patient, taking into account clinical and para-clinical findings at the disease's onset and at subsequently specified points of evolution. The model is applied in three sequential steps. In its first step, it analyses the situation in which a first clinical manifestation suggesting multiple sclerosis occurs - the clinically isolated syndrome (CIS). A key issue for the patient with the CIS is to know the likelihood of having multiple sclerosis. Calculated conditional probabilities at 10 years of evolution, are based on the results of the baseline MRI (normal vs. abnormal) or in the presence vs. absence of oligoclonal bands in the CSF, as presented in Table 3. The conditional probabilities of conversion at 20 years of evolution are calculated based only on the baseline MRI. Considering the same factors, the likelihood that the patients having relapsing-remitting multiple sclerosis (RRMS) will convert into secondary progressive multiple sclerosis (SPMS) at 10 years (Table 4), and an initial estimate for the level of disability of the disease at 20 years (Table 5) are also presented. Results from the BENEFIT study were included to illustrate the effect of disease modifying agents (DMDs) on reducing the 2-year probability of conversion to CDMS in CIS patients with abnormal MRI. In its second step, the model focuses on the clinical status of patients with relapsing-remitting multiple sclerosis (RRMS) in the first year of the disease. Based on the BREMS score, the model calculates the probability of conversion to

CIS-CDMS conversion: source probabilities.

10 Years

20 Years

MRI (O’Riordan et al., 1998) Overall (N =81) Abnormal (N =54) 61.7% 83.3%

Normal (N =27) 18.5%

OCB (Dobson et al., 2013) Overall (N =1759) OCB +(N =1143) 49.57% 64.13%

OCB (N =616) 22.56%

MRI (Fisniku et al., 2008) Overall (N =107) Abnormal (N =73) 62.61% 82.19%

Normal (N =34) 20.59%

Treatment impact – BENEFIT study (Kappos et al., 2006) Interferon beta No treatment 2 years 28% 45% MRI: Magnetic Resonance Imaging; OCB: Oligoclonal bands in the CSF. Please cite this article as: Veloso M. A web-based decision support tool for prognosis simulation in multiple sclerosis. Multiple Sclerosis and Related Disorders (2014), http://dx.doi.org/10.1016/j.msard.2014.04.005

4

M. Veloso

SPMS at 10 years of evolution, as well as the potential benefit of therapy on the probability of conversion into SPMS and on the likelihood of a severe level of disability (EDSS4 = 6) at 20 years of evolution (Table 6). In its third step, the main focus of the model is to assess the evolution of the patient in consecutive years and to predict the level of severity (i.e. low, moderate or severe) for the individual patient at 10 and 20 years of disease evolution. Yearly introduced EDSS will be graphically displayed placing the patient's position within the correspondent MSSS deciles and disability curves. Then, the model will graphically present the anticipated progression of disability curves percentiles (10 years) and MSSS deciles (20 years). By introducing the number/severity of relapses and new lesions on MRI along with the EDSS, the model will present recommendations regarding the possible need for therapeutic scaling, and the probability of conversion to SPMS at 10 and 20 years evolution will be (re)calculated (Table 7).

Table 4

RRMS-SPMS conversion: source probabilities.

Conversion probabilities of RRMS patients into SPMS at 10 years MRI (O’Riordan et al., 1998) Overall (N =52) Abnormal (N =47) 25% 27.66%

Normal (N= 5) 0%

OCB (Amato and Ponziani, 2000) Overall (N =190) OCB + (N =155) 28.94% 34.19%

OCB (N =35) 2.86%

RRMS: Relapsing-Remitting Multiple Sclerosis; SPMS: Relapsing-Remitting Multiple Sclerosis; MRI: Magnetic Resonance Imaging; OCB: Oligoclonal bands in the CSF.

Table 5

Table 8 presents the results of the supplementary model validation tests using the patient dataset of 50 patients. The simulation model (Figure 1) and user instructions are available at http://www.arn.pt/Multiple_Sclerosis/Prognos tic_Models_files/MSprognosisSimulation.html.

4.

Discussion

The current model was designed to answer fundamental questions presented by the patient to the doctor at various evolutionary stages of their disease. For such purpose, it proposes to aggregate, summarize and present relevant scientific evidence in a simple and easy to understand format, through a tool that facilitates a personalized doctor–patient interaction throughout the prognosis discussion.

4.1.

CIS-CDMS conversion

The baseline MRI is the most important prognostic factor to identify which patients with CIS will convert to clinically definite multiple sclerosis (CDMS). MRI allows clinicians to define not only the risk of conversion to CDMS but also correlates with the level of disability at 5 years (Tintore et al., 2006). Such predictive value has been confirmed in 10 and 20 years of evolution (O’Riordan et al., 1998; Fisniku et al., 2008), but in this last study, 18% of patients with abnormal baseline MRI did not develop CDMS. In contrast, a normal baseline MRI is a strong prognostic factor against CDMS conversion (Brex et al., 2002). Nevertheless, some caution about these predictions is required. Baseline MRIs were performed in 0.5 and 1.5 Tesla scanners. We do not know what are the results of different MRI capabilities. The sensitivity for OCBs in the diagnosis of MS was 96,2% and the specificity was 92,5% and increases the positive predictive power of MRI (Villar et al., 2005). In another

Initial estimation for the disability level at 20 years. No lesions: 0 (N =34)

No lesions: 4–9 (N =20)

No lesions: 1–3 (N= 22)

20 Years prediction based on the number of lesions in the baseline MRI (Fisniku et al., 2008) EDSSo =3 25 (75.52%) 14 (63.63%) 10 (50%) 3oEDSSo6 7 (73.52%) 4 (18.18%) 3 (15%) EDSS4 =6 2 (5.88%) 4 (18.18%) 7 (35%)

No lesions: 4= 10 (N =31)

11 (35.43%) 6 (19.35%) 14 (45.16%)

MRI: Magnetic Resonance Imaging; EDSS: Expanded Disability Status Scale.

Table 6

RRMS-SPMS conversion: source probabilities from BREMS (Bergamaschi et al., 2012).

10 Years

20 Years

First quartil

Fourth quartil

First quartil

Fourth quartil

NoTreated

Treated

NoTreated

Treated

NoTreated

Treated

NoTreated

9.1%

1.5%

31.3%

4.1%

26.5% 7% 64.4% Probabilities of reaching EDSS 6 14% 4% 33%

Treated 25.4% 9%

BREMS: Bayesian Risk Estimate for Multiple Sclerosis; RRMS: Relapsing-Remitting Multiple Sclerosis;SPMS: Relapsing-Remitting Multiple Sclerosis. Please cite this article as: Veloso M. A web-based decision support tool for prognosis simulation in multiple sclerosis. Multiple Sclerosis and Related Disorders (2014), http://dx.doi.org/10.1016/j.msard.2014.04.005

Web-based decision support tool in MS

5

Table 7 RRMS-SPMS conversion probabilities based on the age at disease onset, the number of relapses in the first two years and disease duration (Scalfari et al., 2013). Disease duration = 10 years Age at disease onset

Rel y1–2a 1 Attack 2 Attacks 3 Attacks a

Disease duration =20 years Age at disease onset

20 Years

25 Years

30 Years

35 Years

40 Years

20 Years

25 Years

30 Years

35 years

40 Years

6.1 6.8 7.5

7.8 8.6 9.6

10.0 11.9 12.1

12.6 14.0 15.5

16.1 17.9 19.8

12.8 14.1 15.6

16.3 18.0 20.0

20.8 23.0 25.5

26.5 29.3 32.5

33.8 37.4 41.4

Number of relapses in the first 2 years.

Table 8

Model validation: (patient validation group). At 10 years At 20 years (N =50) (N =17)

Model classification of the disability level (% of correct classifications): MSSS decile at year 2 73% 73% MSSS decile at year 5 80% 73% Baseline MRI (no of lesions) – 64% alone Correlations (Pearson coefficient (p-value)): EDSS derived from disability 0.75 curves percentiles2EDSS at (po0.0001) year 10 Patient EDSS in year 22EDSS 0.60 at years 10 (po0.0001) Patient EDSS in year 52EDSS 0.77 at years 10 (po0.0001) Patient MSSS in year 22EDSS 0.58 at years 10 & 20 (po0.0001) Patient MSSS in year 52EDSS 0.77 at years 10 & 20 (po0.0001)

–

– – 0.54 (p =0.0431) 0.56 (p =0.0169)

Patient disability classification using the MS Severity Rank calculator (% of correct classifications): MS Severity Rank calculator – 65% (MS curves) at year 2 MS Severity Rank calculator – 59% (MS curves) at year 5 EDSS: Expanded Disability Status Scale; MSSS: Multiple Sclerosis Severity Score; MRI: Magnetic Resonance Imaging.

study, the detection of OCBs in the CSF indicated a sensitivity of 91,4% and a specificity of 94,1% for a second clinical attack in patients with CIS (Masjuan et al., 2006). A recent meta-analysis confirms that OCB positivity strongly predicts conversion from CIS to MS (Dobson et al., 2013). Taking into account such statistical relevance, the model reproduces these data, as summarized in Table 3. The user must select either one or the other of these factors. When both of these factors are simultaneously instantiated, the model presents the joint probability of the two factors, together with a warning message stating that those specific probabilities have not been validated. Although the combination of these two factors seems to

increase the prognostic power for MS conversion, especially in patients with normal baseline MRI (Tintoré et al., 2008), the fact is that a proper validation is required.

4.2.

RRMS-SPMS conversion

The conversion of relapsing-remitting to secondary progressive multiple sclerosis is analysed by the model in the three stages. In the initial prediction, the baseline MRI and the presence of OCB is the CSF are used by the model to present the probability of conversion to secondary progressive multiple sclerosis (SPMS) of those CIS patients who will convert to relapsing-remitting multiple sclerosis (RRMS). None of the RRMS patients with normal baseline MRI (O’Riordan et al., 1998) and only 2,86% of the patients with negative oligoclonal bands in the CSF (Amato and Ponziani, 2000), did converted to SPMS after 10 years of disease evolution (Table 4). In the second step of the model, the probability of conversion to secondary progressive at 10 years of disease evolution based on the BREMS score calculated in the first year of the disease is presented. The BREMS score for any given patient is calculated from the sum of the relative risk of the prognostic factors included in the score, outlined in Table 1. That relative risk has been derived from a model exploiting the Bayesian methodology and Markov chain Mont Carlo (MCMC) simulation technology, specifying the full joint probability distribution for the variables characterizing the entire course of disease of 186 patients with an initial RRMS (Bergamaschi et al., 2001). The model will present these results only when the BREMS score of the patient is within the first or the fourth quartile, taking into account that only these two quartiles have demonstrated statistical significance. Among the patients with a low BREMS score (1st quartile distribution) only 4% developed SPMS after 10 years of evolution (low risk), whereas patients with high scores (4th quartile) had 29% probability to reach that level (high risk) (Table 6) (Bergamaschi et al., 2012). When evaluating the BREMS score as a diagnostic test to predict the risk of reaching SPMS at 10 years from clinical onset, the authors reported a specificity 0.99 and a sensitivity 0.17, yielding a positive predictive value 0.86 and negative predictive value 0.83 (Bergamaschi et al., 2007). When assessing the disease progression in its third step, the model performs a new calculation of the conversion probabilities to SPMS (Table 7) at 10 and 20 years of disease

Please cite this article as: Veloso M. A web-based decision support tool for prognosis simulation in multiple sclerosis. Multiple Sclerosis and Related Disorders (2014), http://dx.doi.org/10.1016/j.msard.2014.04.005

6

M. Veloso

Fig. 1

Model interface.

evolution, based on the age at disease onset and the number of relapses in the first two years (Scalfari et al., 2013). This focus of the model in the secondary progressive phase, presenting three distinct estimations depending on the context, results from the fact that the onset of the SP phase is a robust marker for late disability (Scalfari et al., 2013), and that it appears to be a good outcome measure with relatively high inter-rate reproducibility (Minderhoud et al., 1988). Patients with a poor prognosis have a shorter relapsing-remitting phase and, as such, enter the secondary progressive phase at a younger age (Kremenchutzky et al., 2006). At the same time, the influence of clinical factors in the disability progression seems to be confined to the period between disease's onset and attainment of EDSS 4 (Confavreux et al., 2003; Debouverie et al., 2008).

Moreover, in about 85% of the cases, the beginning of the progressive phase occurs with this EDSS (Tremlett et al., 2008).

4.3.

Estimation of the disability level

In its first stage, the model proposes an initial estimation of the anticipated level of disability at 20 years of disease evolution (Table 5). Although, in the long term, the number of lesions of the baseline MRI has limited predictive value (Fisniku et al., 2008), such information may be valuable to the clinician to adjust the initial prognosis discussion with the individual patient with CIS. Nevertheless, 64% accuracy was obtained in the tests using the validation group of 50

Please cite this article as: Veloso M. A web-based decision support tool for prognosis simulation in multiple sclerosis. Multiple Sclerosis and Related Disorders (2014), http://dx.doi.org/10.1016/j.msard.2014.04.005

Web-based decision support tool in MS patients (Table 8), but all included patients have RRMS, not CIS. The patient's position within the MSSS deciles, derived from the EDSS in each year of follow up specified has been considered particularly important. The Multiple Sclerosis Severity Score (MSSS) is defined as the median decile rank of each EDSS value in the population of MS patients with similar disease duration. The global MSSS table was derived from the combined data on 9892 untreated patients from 11 countries. By ranking the EDSS scores, MSSS can be invoked as a linear measure, as opposed to EDSS (Roxburgh et al., 2005). Therefore, MSSS provides the ability to assess and compare disease severity in MS patients at all EDSS levels for any given disease duration. The likelihood that the patient will remain in the same decile, especially after the 5th year of evolution, is high (Gray et al., 2008). Similarly, the validation of the constructed disability curves demonstrated that the probability of deviating from the initially assigned percentile to a higher percentile over time, was in the range of 6,5% for the 50th percentile to 10.4% for the 75th percentile (Achiron et al., 2003). As to the model 10 years predictions, a strong correlation was found with the disability curves (Pearson coefficient= 0.75; po0.0001), and 73% probabilities of a correct classification of the estimated level of disability at 20 years of disease evolution, using the MSSS projections, was obtained (Table 8). Furthermore, the simulation model performed slightly better than the MS curves severity rank calculator in the long term severity classification tests. The reduced number of patients in this validation group yields limited statistical value, but it seems to be in line with the published results. Along with the representation of disease progression through the MSSS deciles and disability curves percentiles, the model shows a probability distribution for the three levels of disability that represents a kind of confidence on the prediction derived from the MSSS decile regarding each disability level.

4.4.

Assessment of treatment impact

There is widespread consensus showing that treatment with DMDs reduces the risk of further attacks, lowers the MRI burden, and slows progression of neurologic disability. For instances, the BENEFIT study demonstrated that, after 2 years, treatment with interferon beta in patients with CIS significantly decreased the conversion to CDMS from 45% of the placebo group to 28% (Kappos et al., 2006). Moreover, the 5-year open label active treatment extension of the BENEFIT trial confirmed that the risk for conversion to CDMS was significantly lower in the early treatment group than in the delay treatment group (Kappos et al., 2009). Long-term efficacy of DMDs is also provided for RRMS. The model shows how immunomodulatory treatment affects the probabilities of secondary progressive conversion at 10 and 20 years, as well as to achieve EDSS4 = 6 at 20 years, depending on the BREMS score in the first year of the disease (Table 6). The model uses data from the BREMS study, which concludes that DMDs significantly reduce the risk of progression of multiple sclerosis in patients with low and high BREMS scores (Bergamaschi et al., 2012). Clearly one should face this beneficial effect with some caution.

7 Information about the long term benefits of therapy on disease progression is scarce and conflicting (Pachner and Steiner, 2009; Shirani et al., 2012; Tedeholm et al., 2013). A modification of the MSSS decile or the disability curves percentile should alert the clinician to a potential need for therapeutic change. The therapeutic recommendations presented by the model are based on widely accepted criteria (Zaffaroni, 2005; Freedman et al., 2004), but additional warning information is provided if, in spite of a criterion proposing a “low level of concern” message, a severe disability is predicted according to the current MSSS decile.

4.5. Comment, limitations and future developments Various other prognostic factors have been described, such as a short interval between the initial event and the second relapse (Tremlett et al. 2010; Confavreux et al., 2003). Other factors like gender and the involvement of various functional systems in the initial episode are already encompassed in the calculation of the BREMS score. Similarly, the level of disability in the first five years of evolution, another described prognostic factor, is incorporated into the model through the MSSS representation. However, an important limitation of the current model, that should be considered in future model developments, is the absence of further disability dimensions, namely regarding the cognitive deficits (Chiaravalloti and DeLuca, 2008). The lack of long term consistent data on this subject contributed to this limitation. EDSS and the SP phase are by far the outcomes measures more often used in the long term/natural history studies. Another limitation of this model is about the therapeutic assessment. First, the assessment of the potential treatment benefit is carried out exclusively with data from the patient's 1st year of the disease. The lack of data does not enable us to update the therapeutic effect accordingly on the subsequent clinical course of the patient. Therefore, this prediction must be interpreted exclusively in that context (1st year of disease), actually accordingly to the design of the model layout and, in a way, underscoring that greater benefits on disability may be obtained by an early interferon beta treatment in RRMS, as also demonstrated in large observational studies (Trojano et al., 2009). Second, only the immunomodulatory therapeutic has been contemplated. Finally, the model accuracy regarding long term prediction of the disability level is far from optimal, but this reflects existing scientific evidence on this subject. Difficulties of individualized prognosis counselling in multiple sclerosis are widely recognized, which favours the utilization of decision support tools to overcome existing difficulties, by at least to disseminating the current scientific evidence. One such tool is “Evidence-Based Decision Support tool in Multiple Sclerosis” (EBDiMS) in which the software performs an online search for the best matching patients, of a reference database, that most closely matches the individual characteristics of the patient such as, disease course, age at first symptoms, attack number in the first two years, first inter-interval attack and/or time to EDSS3, and it calculates prognoses such as time to

Please cite this article as: Veloso M. A web-based decision support tool for prognosis simulation in multiple sclerosis. Multiple Sclerosis and Related Disorders (2014), http://dx.doi.org/10.1016/j.msard.2014.04.005

8

M. Veloso

conversion to SPMS or time to reach EDSS 6/8/10. As demonstrated in the evaluation of EBDiMS, its predictive accuracy, as compared with of 17 neurologists highly specialized in MS, was similar and more consistent (Galea et al., 2013). Another evaluation of the online tool found that clinicians and patients were only moderately interested in the tool, and that its use did not change the prognostic estimate previously made by patients (Heesen et al., 2013). These conclusions reinforce the idea that is not enough to have an accurate and sound model. These kinds of tools should satisfy the needs of users and present an appealing and easy to use interface. The current model utilizes a different approach. It attempts to simulate patient prognoses at fundamental phases of MS evolution in a practice oriented way. Additionally, the model presents prognoses as probabilities to reach outcomes at 10 and 20 years, instead of time to reach them. Ultimately, the model proposes conveying the research results of relevant studies regarding MS prognosis in a comprehensive and easy to use format. The selected prognostic factors are widely used in clinical practice and were obtained from studies based on large patient datasets reporting significative/validated data. These characteristics provide the model for the scientific evidence absolutely necessary to accredit its use in the clinical practice, whether accepted by clinicians. Finally, as highlighted in the evaluation of EBDiMS, the psychological impact that such instruments have on patients cannot be ignored, leaving the clinician to decide on the most appropriate kind of prognostic discussion for each individual patient.

Conflict of interest The author has no conflicts of interest or disclosure to report. This work has received no financial support.

References Achiron A, Barak Y, Rotstein Z. Longitudinal disability curves for predicting the course of relapsing-remitting multiple sclerosis. Mult Scler 2003;9:486–91. Amato MP, Ponziani G. A prospective study on the prognosis of multiple sclerosis. Neurol Sci 2000;21:S831–8. Bergamaschi R, Berzuini C, Romani A, Cosi V. Predicting secondery progression in relapse-remitting multiple sclerosis: a Bayesian analysis. J Neurol Sci 2001;189:13–21. Bergamaschi R, Quaglini S, Trojano M, Amato MP, Tavazzi E, Paolicelli D, et al. Early prediction of the long term evolution of multiple sclerosis: the Bayesian Risk Estimate for Multiple Sclerosis (BREMS) score. J Neurol Neurosurg Psychiatry 2007;78:757–9. Bergamaschi R, Quaglini S, Tavazzi E, Amato MP, Paolicelli D, Zipoli V, et al. Immunomodulatory therapies delay disease progression in multiple sclerosis. Mult Scler J 2012;0(0):1–9. Brex PA, Ciccarelli O, O’Riordan JI, Sailer M, Thompson AJ, Miller DH. A longitudinal study of abnormalities on MRI and disability from multiple sclerosis. N Engl J Med 2002;346:158–64. Chiaravalloti ND, DeLuca J. Cognitive impairment in multiple sclerosis. Lancet Neurol 2008;7(12):1139–51. Confavreux C, Vukusic S, Adeleine P. Early clinical predictors and progression of irreversible disability in multiple sclerosis: an amnesic process. Brain 2003;126:770–82.

Debouverie M, Pittion-Vouyovitch S, Luis S, Guilemin F. Natural history of multiple sclerosis in a population-based cohort. Eur J Neurol 2008;15:916–21. Dobson R, Ramagopalan S, Davis A, Giovannoni G. Cerebrospinal fluid oligoclonal bands in multiple sclerosis and clinically isolated syndromes: a meta-analysis of prevalence, prognosis and effect of latitude. J Neurol Neurosurg Psychiatry 2013;0:1–6. Dong Y, Chbat NW, Gupta A, Hadzikadic M, Gajic O. Systems modeling and simulation applications for critical care medicine. Ann Intensive Care 2012;2:18. Fisniku LK, Brex PA, Altmann DR, Miszkiel KA, Benton CE, Lanyon R, et al. Disability and T2 MRI lesions: a 20-year follow-up of patients with relapse onset of multiple sclerosis. Brain 2008;131:808–17. Freedman MS, Patry DG, Grand’Maison F, Myles ML, Paty DW, Selchen DH, et al. Treatment optimization in multiple sclerosis. Can J Neurol Sci 2004;31:157–68. Galea I, Lederer C, Neuhaus A, Muraro PA, Scalfari A, Koch-Henriksen N, et al. A Web-based tool for personalized prediction of long-term disease course in patients with multiple sclerosis. Eur J Neurol 2013;20:1107–9. Gray OM, Jolley D, Zwanikken C, Trojano M, Grand’Maison F, Duquette P, et al. The Multiple Sclerosis Severity Score (MSSS) re-examined: EDSS rank stability in the MSBase dataset increases 5 years after onset of multiple sclerosis. Neurol Asia 2008;13:217–9. Heesen C, Solari A, Giordano A, Kasper J, Kö pke S. Decisions on multiple sclerosis immunotherapy: new treatment complexities urge patient engagement. J Neurol Sci 2011;306(1-2):192–7. Heesen C, Gaissmaier W, Nguyen F, Stellmann J-P, Kasper J, Köpke S, et al. Prognostic risk estimates of patients with multiple sclerosis and their physicians: comparison to an online analytical risk counseling tool. PLoS One 2013;8(5): e59042. http://dx.doi.org/10.1371/journal.pone.0059042. Hughes SE, Spelman T, Bianchi E, Méchattii S, Butzkueven H. For the MSBase investigators. MS Curves: a new online tool for assessing MS severity using the MSBase Registry. In: Proceedings of the ECTRIMS 28th conference of the European committee for the treatment and research in multiple sclerosis. Lyon, France; 10– 13 October 2012. Also available at 〈http://www.msbase.org〉. Kister I, Chamot E, Salter AR, Cutter GR, Bacon TE, Herbert J. Disability in multiple sclerosis: a reference for patients and clinicians. Neurology 2013;80:1018–24. Kappos L, Polman CH, Freedman MS, et al. for the BENEFIT Study Group. Treatment with interferon beta-1b delays conversion to clinically definite and McDonald MS in patients with clinically isolated syndromes. Neurology 2006;67:1242–9. Kappos L, Freedman MS, Polman CH, Edan G, Hartung HP, Miller DH, et al. Long term effect of early treatment with interferon beta1b after a first clinical event suggestive of multiple sclerosis: 5year active extension of the phase 3 BENEFIT trial. Lancet Neurol 2009;8:987–97. Kremenchutzky M, Rice GP, Baskerville J, Wingerchuck DM, Ebers GC. The natural history of multiple sclerosis: a geographically based study 9: observations on the progressive phase of the disease. Brain 2006;129:584–94. Masjuan J, Alvarez-Cermeno JC, Garcia-Barragan N, Díaz-Sanchez M, Espiño, Sádaba MC, et al. A new oligoclonal band test accurately predicts conversion to MS. Neurology 2006;66:576–8. Minderhoud JM, van der Hoeven JH, Prange AJ. Course and prognosis of chronic progressive multiple sclerosis. Results of an epidemiological study. Acta Neurol Scand 1988;78:10–5. O’Riordan JI, Thompson AJ, Kingsley DP, MacManus DG, Kendall BE, Rudge P, et al. The prognostic value of brain MRI in clinically isolated syndromes of the CNS. A 10-year follow-up. Brain 1998;121(Pt 3):495–503. Pachner AR, Steiner I. The multiple sclerosis severity score (MSSS) predicts disease severity over time. J Neurol Sci 2009;278:66–70. Roxburgh RHSR, Seaman SR, Masterman T, Hensiek AE, Sawcer SJ, Vukusic S, et al. Multiple Sclerosis Severity Score: using disability

Please cite this article as: Veloso M. A web-based decision support tool for prognosis simulation in multiple sclerosis. Multiple Sclerosis and Related Disorders (2014), http://dx.doi.org/10.1016/j.msard.2014.04.005

Web-based decision support tool in MS and disease duration to rate disease severity. Neurology 2005;64:1144–51. Scalfari A, Neuhaus A, Daumer M, Muraro PA, Ebers GC. Onset of secondary progressive phase and long-term evolution of multiple sclerosis. J Neurol Neurosurg Psychiatry 2013;00:1–9. Shirani A, Zhao Y, Karim ME, Evans C, Kingwell E, van der Kop ML, et al. Association between use of interferon beta and progression of disability in patients with relapsing-remitting multiple sclerosis. J Am Med Assoc 2012;308:247–56. Tedeholm H, Lycke J, Skoog B, Lisovskaja V, Hillert J, Dahle C, et al. Time to secondary progression in patients with multiple sclerosis treated with first generation immunomodulating drugs and historical controls, adjusted for clinical covariates. Mult Scler J 2013;19(6):765–74. Tintore M, Rovira A, Rio J, Nos C, Grivé E, Tellez N, et al. Baseline MRI predicts future attacks and disability in clinically isolated syndromes. Neurology 2006;67:968–72. Tintoré M, Rovira A, Rio J, Tur C, Pelayo R, Nos C, et al. Do oligoclonal bands add information to MRI in first attacks of multiple sclerosis? Neurology 2008;70:1079–83. Tremlett H, Zhao Y, Devonshire V. Natural history of secondaryprogressive multiple sclerosis. Mult Scler 2008;14:314–24. Tremlett H, Zhao Y, Rieckmann P, Hutchinson M. New perspectives in the natural history of multiple sclerosis. Neurology 2010;74: 2004–15.

9 Trojano M, Liguori M, Bosco Zimatore G, Bugarini R, Avolio C, Paolicelli D, et al. Age-related disability in multiple sclerosis. Ann Neurol 2002;51(4):475–80. Trojano M, Pellegrini F, Paolicelli D, Fuiani A, Zimatore GB, Tortorella C, et al. Real-life impact of early interferon beta therapy in relapsing multiple sclerosis. Ann Neurol 2009;66:513–20. Veloso M. An agent-based simulation model for informed shared decision making in multiple sclerosis. Mult Scler Relat Disord 2013;2:377–84. Villar LM, Masjuan J, Sadaba MC, González-Porqué P, Plaza J, Bootello A, et al. Early differential diagnosis of multiple sclerosis using a new oligoclonal band test. Arch Neurol. 2005;62:574–7. Vukusic S, Confavreux C. Natural history of multiple sclerosis: risk factors and prognostic indicators. Curr Opin Neurol 2007;20:269–74. Zaffaroni M. Treatment optimization in multiple sclerosis. Neurol Sci 2005;26:S187–92. Wyatt JC, Altman DG. Commentary prognostic models: clinically useful or quickly forgotten. Br Med J 1995;311:1539. Wilensky U. NetLogo.Evanston IL: Center for Connected Learning and Computer Based Modeling: Northwestern University; 1999. 〈http://ccl.northwestern.edu/netlogo/〉.

Please cite this article as: Veloso M. A web-based decision support tool for prognosis simulation in multiple sclerosis. Multiple Sclerosis and Related Disorders (2014), http://dx.doi.org/10.1016/j.msard.2014.04.005