Simply asked, the question for this post is “if you have active relapsing-remitting MS will disease modifying therapy mitigate your risk for long-term disability?” The simple answer for the early relapsing-remitting patient is that disease modifying therapy prevents relapses which can result in cumulative and permanent disability. The more complicated question is whether disease modifying therapy decreases long term (20– 30 years) disability. This question is less easily studied than short term outcomes and the answer is less clear for multiple reasons but mainly because the tools that are available to answer this question are problematical.
Active relapsing-remitting MS refers to patients with relapsing-remitting MS who have had relapses within the past 2 years and have new MS lesions on recent MRI scans. Disease modifying therapy has not been studied in inactive MS so that we know little or nothing about the long-term benefits of disease modifying therapy for inactive relapsing-remitting MS patients. Since FDA approval for the use of disease modifying therapy in active secondary progressive MS (SPMS) and primary progressive MS has only recently been granted, we also don’t have long-term follow-up of these patients.
The current model of relapsing remitting MS is one of an initial inflammatory phase followed by a chronic (for lack of a better term) degenerative phase. Short term clinical trials demonstrate that disease modifying therapy reduces active inflammation in the early phase of MS. We would like to know that this effect is translated into improved long-term outcome.
All neurologists agree that disease modifying therapies for active relapsing-remitting MS are effective in the short term and most agree that a few disease modifying therapies may slow worsening in progressive forms of MS (also in the short run), but given the risks of disease modifying therapy, we would also like to be sure that disease modifying therapies improve the long-term outcome in MS.
This post will describe common clinical outcome measures used to measure the long-term outcome in relapsing-remitting MS and then discuss some of the difficulties in answering this question.
Outcome measures in relapsing-remitting MS
The main clinical outcome measures to look at long term outcomes in relapsing-remitting MS are change in the Kurtzke Expanded Disability Status Scale (EDSS) and the time to conversion to secondary progressive MS. With regards to the EDSS, there are four major disability “milestones” for patients with MS:
EDSS Score | Milestone |
3 | When a person first develops moderate disability in one functional system but remains is able to walk without a problem |
4 | Severe disability in one functional system but remains self-sufficient even though there may be some limitation of ambulation |
6 | Need to use an adaptive aid such as a cane or walker to walk |
7 | When they become wheelchair-bound |
Reaching an EDSS of 4 is particularly important as this frequently marks the time when relapsing-remitting MS converts to secondary progressive MS. Please see http://jonathanharrismd.com/outcomes-in-ms/ for more information about the EDSS.
The time to conversion to secondary progressive MS is similarly important as it indicates a change to a different usually more rapid course of progressive worsening.
Secondary progressive MS, in the past, has been diagnosed in retrospect based on a history of gradual neurological worsening after an initial relapsing-remitting course. That is, at a certain time in the course of relapsing-remitting MS, the neurologist notes that a person has begun to progressively worsen with or without relapses. Usually, the diagnosis is made month to years after the change has occurred, as it may take this long to be sure that a permanent change has occurred and that relapses are not still a major feature of the person’s illness.
The lack of a specific point in time for the diagnosis of secondary progressive MS can be a problem. When looking at studies of the natural history of MS, this contributes to a wide range in the times that relapsing-remitting MS patients convert to secondary progressive MS. Also, different studies may use different definitions of secondary progressive MS making it difficult to compare this outcome between studies.
Lorsheider et al (Lorscheider, Buzzard, & Jokubaitis, 2016) used the MSbase registry to test different possible definitions for secondary progressive MS to find a single objective definition that could diagnose secondary progressive MS most reliably at a single point in time. The definition they determined is not quite as accurate as a physician’s retrospective diagnosis of secondary progressive MS but provides an objective definition for secondary progressive MS and allows the diagnosis of secondary progressive MS more than 3 years earlier than a retrospective diagnosis.
Minimum EDSS of 4 | Moderate disability in one functional system / fully ambulatory, self-sufficient, up 12 hours a day despite relatively severe disability. Able to walk ≥500 meters without aid/rest | |
Minimum pyramidal functional system score of 2 | Some weakness in one or more limbs causing at least minimal disability | |
Disability progression in the absence of a relapse | One step on the EDSS in patients with an EDSS of less than or equal to 5 | One-half step in a patient with an EDSS of greater than 6 |
Continued progression over greater than or equal to 3 months including confirmation within the leading functional scale |
Why is it difficult to determine to determine the long-term efficacy of disease modifying therapy?
The primary reason is that MS is a disease that progresses over decades. Decades long randomized placebo-controlled trials would be necessary to definitively establish the long-term efficacy of disease modifying therapy. Randomized placebo-controlled trials paid for by pharmaceutical companies to obtain Food and Drug Administration (FDA) approval to sell their drug are too short to establish long-term efficacy.
Long-term studies are not feasible partly for ethical reasons and partly because of their expense. From an ethical standpoint, since medications are available which decrease relapse rate and short-term disability, it is unethical to have a control group of MS patients that are not treated with a disease modifying therapy. Additionally, given the availability of these drugs, patients are unlikely to agree to remain without treatment and the neurologist treating them would discourage it. (Goodin, Traboulsee, Knappertz, & et al., 2012).
From a cost standpoint, even short clinical trials are expensive. These costs are paid by pharmaceutical companies whose exclusive patent rights to market a drug last only 20 years from the date of patent application. The 20-year “clock” starts as soon as initial patent rights are granted. Then, it may take several years before obtaining FDA approval to market the drug. An expensive, decades long clinical trial, might not end until after a drug loses its patent. This would make it impossible for a pharmaceutical company to recoup its expenses. Additionally, other newer and better drugs are likely to be discovered while doing a long clinical trial, making the long-term effect of an older drug less important.
Without long-term randomized clinical trials, what is available to show the long-term efficacy of disease modifying therapy?
To help answer this question, we are left with deductive reasoning, and observational studies using patient registries in one form or another.
Deductive reasoning
Deductive reasoning is a method of reasoning by which premises understood to be true produce logically certain conclusions (Contributors, 2020). Some neurologists use deductive reasoning without definite clinical evidence to approach the question of long-term benefits of disease modifying therapy. They argue that: (1) natural history studies suggest relapses early in the course of relapsing remitting multiple sclerosis impact early progression of disability and (2) that increases in the volume of lesions on MRI early in the course of relapse remitting disease correlates with degree of long term disability (Multiple Sclerosis Coalition, 2019); so that it makes sense (though does not prove) that disease modifying therapies that will decrease relapses and the volume of MRI lesions will decrease long term disability. Also, given the current concept of multiple sclerosis which involves the interaction between inflammation and degeneration of the central nervous system, medications which treat inflammation should alter the long-term outcome of the disease. Although these deductions seem valid, their conclusions are not necessarily true.
Patent Registries
A patient registry is a system that uses observational study methods to collect data to evaluate outcomes for a population with a particular disease. (Gliklich, Dreyer, & M, 2020) Patient registries offer the ability to evaluate patient outcomes when clinical trials are not practical or ethically acceptable, or over very long-time periods. (Gliklich, Dreyer, & M, 2020)
Disease registries follow patients who have their medical records entered into data banks of various types and consent to have their records followed over long periods of time. In MS, they are used for natural history studies of untreated patients, open label extension trials, and to compare outcomes in patients treated on disease modifying therapy. Older studies compare patients treated on disease modifying therapy to “historical” patients (using medical records of patients before disease modifying therapy was available) and to contemporary patients not treated on disease modifying therapy. Most recently, registries have been used to compare contemporary patients treated with different disease modifying therapies to each other and to untreated patients.
There are several steps one needs to follow when analyzing and interpreting studies using registry data. According to Gliklich et al. (Gliklich, Dreyer, & M, 2020), one needs to ask the following 6 questions:
The MSBase Registry is an example of a large MS registry. This is an international collaboration and consists of the largest organized repository of longitudinal, “real-world” MS patient data. The Registry commenced in 2004 and has accumulated over 52,000 patient records from 33 participating countries. Other large-scale registries include the Danish MS Registry, the Norwegian MS Registry, the Swedish MS Registry, the Italian MS Database Network, the North-American NARCOMS Registry, and the German MS Registry ( MS International Federation, 2018).
Registries rarely include everyone with a certain disease from a circumscribed geographic area. Nevertheless, only this type of inclusive registry can be representative of a general population of MS patients and used for a true natural history study. Of the registries listed above, the Danish MS Registry is different from the others as it was established in 1956 and contains data on all Danes who have been diagnosed after 1921 and who were alive in 1948 or have been diagnosed and have been reported since then in Denmark. The other registries do not contain all of the MS patients from a designated geographical area.
Two features need to be kept in mind when looking at long term outcome studies using patients from MS registries. First, who are the patients entered into the registry? MS centers contributing to registries are frequently tertiary referral MS centers and as such, their patient population is enriched by patients with active MS and with those patients treated with immunomodulatory agents. Early adopters are physicians who are quick to use a new drug when it comes to market. Their practice differs from physicians who prefer to use a drug after it is well established. Early adopters such as academic physicians may be more like to enter their patients into a registry which is another reason why MS registries may not be representative of the general population.
The MSbase registry, for example, enters patients from tertiary MS centers and presumably physicians who are early adopters. Thus, patients in this registry may not be representative of the MS population in general.
The second feature to be kept in mind is that studies using most MS registries have objectives and hypothesis that are determined after the registry was set up. Data already collected in the registry is not collected specifically to answer the question(s) of the study.
Bias and confounding
Two major problems occur when randomized clinical trials cannot be done to answer a clinical question. These problems are bias and confounding. Bias is defined as any systematic error in an epidemiological study that results in an incorrect estimate of the true effect of an exposure on the outcome of interest (Barratt, Kirwan, & Shantikumar, 2020). The definition of the term confounding can be confusing. Formally, confounding occurs when an observed association between an exposure to an independent variable and a dependent variable is distorted or spurious because the exposure to the independent variable is also correlated with another factor (a “confounder”) that affects the dependent variable (outcome) irrespective of the effects of the independent variable (Barratt, Kirwan, & Shantikumar, 2020). Informally, the term confounding is also used to describe a situation where an independent variable, different and unrelated to the chosen independent variable has an effect on the (chosen) dependent variable. With both the formal and informal use of the term, the outcome in question may be totally or partially due to the confounder and not the chosen independent variable.
To clarify this concept, the independent variable for the purpose of this post is the use of a disease modifying therapy and the dependent variable is long term MS outcome. Age and socioeconomic status are examples of two common independent variables that can affect long term MS outcomes, irrespective of whether disease modifying therapy is used. Both can also affect the disease modifying therapy a physician chooses. As such, age and socioeconomic status are “formal” confounders. In a country with a national health care plan, socioeconomic status could be a factor which affects MS outcomes but may not affect choice of disease modifying therapy and thus would be an informal confounder. When looking at the effect of disease modifying therapy on long term MS outcome, one must consider the potential “confounding” effect of age and socioeconomic status before concluding that disease modifying therapy is responsible for the patient’s outcome.
Confounders may be known or unknown. In a randomized clinical trial, one expects the effect of confounders to be the same in each randomly selected group so that confounders should not unduly influence outcome. In non-randomized groups, known and unknown confounders may be more common in one group than another. This would make confounding factors more likely to influence the outcome in one group than another. Known confounders can be controlled for in a study; unknown confounders can’t. Unknown confounders can only be controlled by using groups chosen at random from the same population.
The following tables note common types of bias and confounders that one needs to watch for in outcome studies.
Table 1
Bias | Systemic error that results from factors that relate both to the specific treatment chosen and an outcome. |
Attrition bias (Loss to follow-up bias) | Loss of participants to follow-up. |
Censored patients | Patients enrolled early in a study have greater likelihood of an event occurring than those who enter later. Censored patients are those in whom the event did not occur before the study ended. |
Confirmation bias | Tendency to look for evidence that will confirm a hypothesis and failing to look for other explanations |
Exclusion bias | Systematic exclusion of certain individuals from a study. |
Immortal time bias | Immortal time refers to a span of time in the observation or follow-up period of a cohort during which the outcome under study could not have occurred. Generally, this is the time before the exposure to a treatment when the treatment can’t have any effect on the outcome. This bias systematically overestimates the outcome rate in the unexposed group and at times also underestimates the rate in the group exposed to the medication. As a result, the rate ratio of exposure is underestimated, creating the illusion that the drug is effective in preventing the outcome under study. (Suissa, 2008) |
Misclassification in treatment (Form of misclassification bias) | Patient incorrectly categorized with respect to their exposure to therapy |
Recall bias | The patient has an incorrect recollection of the dose of medication or adherence to a treatment |
Sample selection bias | Samples that are collected to determine their distribution are selected incorrectly and do not represent the true distribution because of nonrandom reasons (scholarpedia.org) |
Selection bias | Selection of a patient for study or the likelihood of their being retained in the study leads to a different result than you would have gotten if you had enrolled the entire target population (sphweb.bumc.bu.edu) |
Self-selection bias | When participants choose whether or not to participate in a project the group that chooses is not equivalent to the group that ops out. Occurs when people volunteer for a study. The people that get a medication differ from those who don’t in “goodness knows” how many ways (Nisbett, 2015) |
Selective loss to follow-up | Sickest patients or those with outcomes of greatest interest are lost to follow-up |
Status quo bias | People prefer things to stay the same by doing nothing, or by sticking to a decision previously made (Behavioral Science Solutions Ltd, 2020) |
Survivorship bias | Error from concentrating on the people that make it past some selection process and overlooking those that did not (Contributors, Wikipedia, 2020). |
Table 2
Confounders | Factors which influence a treatment selection and also affect an outcome irrespective of the treatment chosen. (Gliklich, Dreyer, & M, 2020). | ||
Known | |||
Patient related | Age, gender, race, socioeconomic class, disease severity, lifestyle characteristics and comorbid illnesses | ||
People who receive a new treatment have different risk factors for adverse events than those who choose other treatment or no treatment | |||
Confounding by indication or “selective prescribing” – When people with more severe disease or who have failed other therapies are more likely to receive newer treatments. These patients are systematically different from other patients who are treated with medication. | |||
Physician related | Experience and skills | ||
Early adopters – Their practice differs from physicians who prefer to use a drug after it is well established. | |||
System factors | Type of care settings, quality of care, and regional effects | ||
Unknown |
Of particular importance to long term studies of disease modifying therapy are confounders related to: (1) why each patient was selected to be on a particular disease modifying therapy or left untreated; (2) when treatment began with regard to the course of their disease; and (3) the fact that neither patient nor physician is blinded to the treatment received.
Multiple regression analysis
Multiple regression analysis is commonly used to compare different groups in in observational registry studies. This technique correlates many independent variables simultaneously with a chosen dependent variable. This allows the investigator to isolate the effect of the independent variable he chooses on the dependent variable. In our case., when looking at the dependent variable of outcome with the use of disease modifying therapy in MS, some examples of independent variables are age, sex, length of time diagnosed with MS, level of disability, as well as disease modifying therapy use.
In theory, multiple regression analysis controls for everything that is related to the independent and dependent variables but in practice, it is impossible to identify all possible confounders (Nisbett, 2015). This is due to a problem with the nonrandom selection of patients in groups that are compared. The patients in each group may differ in any number of ways.
A second problem with multiple regression analysis is that even when we are aware of confounders, we may not measure each of the confounding variables well (Nisbett, 2015). According to Nisbett (Nisbett, 2015), socioeconomic class is typically the biggest confounder in epidemiological studies. Richer people can afford better food and better doctors, receive better health care in general, are likely to be better read, and overall to have better life outcomes. If socioeconomic status is considered as a confounder, how it is measured is important to how it effects the dependent variable.
Causality determination in multiple regression analysis can be improved by looking at lagged correlations (outcomes later in time), the magnitude of the effect, and a dose dependent effect (Nisbett, 2015).
Common problems encountered in using observational registry studies to determine the effect of disease modifying therapy on long term MS outcome.
Open label extension studies
Open label extension studies are one type of registry study that has been done after randomized clinical trials of many currently FDA approved medications. In extension studies, people who have participated in a randomized placebo-controlled trial are asked to continue the medication tested after the original trial is complete and enter a registry to allow long term follow-up of their medical condition.
Fingolimod is an FDA approved disease modifying therapy for relapsing-remitting MS. The LONGTERMS study (Cohen, Tenenbaum, Bhatt, Zhang, & Kappos, 2019) is an example of an open label extension study of fingolimod. This study enrolled 4086 relapsing-remitting MS patients aged 18 years or older who had completed earlier studies of fingolimod and continued fingolimod 0.5mg orally once daily. The study followed the frequency of relapses, EDSS, and MRI in these patients and documented medication side effects. Of the 4086 patients initially enrolled, 3480 (85%) completed the study. The median age of participants was 38 with a range from 17–65 years. The median time people took fingolimod was about 2 1/2 years (945 days with a range of 75–4777 days).
The average relapse rate per year was 0.22 in the first two years of this extension study but then decreased to 0.17 over years 0-10. 46% of patients remained relapse free after 10 years and 63% of patients were free from 6-month confirmed disability worsening.
As a group, disability worsened slightly over the course of the study. Average worsening in EDSS from entry into the registry to year 10 was 0.40. By the end of year 10, 68% of the patients did not reach the EDSS milestones of severe disability in one functional system (EDSS⩾4.0), 85% did not require a cane or walker to ambulate (EDS⩾6.0, and 96% were not wheelchair bound (EDSS ⩾7.0).
The number of new/newly enlarging T2 lesions decreased each year after the second year of the study. The mean total volume of T2 lesions increased at years 3 and 4 compared with baseline, and remained stable at later visits. From baseline to the end of the study, mean T2 lesion volume worsened 1589mm3 and there was a 3% average decrease in brain volume.
13% of patients had serious side effects. The most frequently reported side effects were basal cell carcinoma and MS relapse which each occurred in 1% of the patients.
Open label extension studies, like this one, purport to show long-term efficacy and safety of medications. However, the results of these studies are difficult to interpret mainly due to the problem of survivorship and self-selection bias. Survivorship bias is an error from studying only the people that make it past some selection process and overlooking those that did not (Contributors, Wikipedia, 2020). Patients who stay on a specific disease modifying therapy long term are, a priori, those who are doing well on that therapy. The positive outcome may be due to the medication but also could be due to less aggressive underlying MS. Patients who do not do well are likely to discontinue the medication and try a different one. This means that patients continuing to receive treatment in open label extension studies are already selected for a positive outcome before starting in the extension study.
Self-selection bias refers to when participants themselves choose whether to participate in a project. The group that ops in is not equivalent to the group that ops out. In extension studies, the people that continue the medication can differ from those who do not in many ways. Survivorship and self-selection bias prevent long-term extension studies from proving long-term benefit of a disease modifying therapy.
Another problem with open label extension studies is the lack of a comparison group. Published studies generally show good looking data and long-term outcomes that seem better than one would expect from no treatment. But to whom are we comparing these patients? At best, we are comparing these patients to historical controls from the literature and at worst to imaginary patients in our minds rather than actual patients that have been studied in tandem with those in the open label extension trial.
Moreover, in studies like the LONGTERMS study, we do not know what happens to people in the initial trials who did not remain on the medication studied or did not enter the extension study so that we do not know if they did just as well in the long term.
Open label extension studies seem to be reassuring with regards to the long-term risk of side effects from a disease modifying therapy, although once again, one must keep in mind that those with early side effects from the disease modifying therapy would not continue the drug and enter the open label extension study.
Does the LONGTERMS study of fingolimod show an improvement in the long-term outcome of people with active relapsing-remitting MS? The reader needs to keep survivorship and self-selection bias in mind when answering this question to decide if patients who stayed on this medication had long-term benefit with an acceptable risk of side effects.
Comparing current patients with patients from the past
One might ask why not just compare how patients on disease modifying therapy do now compared to how patients did before disease modifying therapy was available? One such study was done in Sweden in 2012 (Tedeholm et al, 2012). In this retrospective study, 186 relapsing remitting MS patients seen at one center in Sweden from 1950-1964 were compared to 730 more contemporary patients from multiple centers in Sweden who were treated with interferon or glatiramer acetate from 1995-2004. The contemporary group treated with interferon or glatiramer took a longer time to reach secondary progressive MS than the historical group. The study concluded that patients treated on the early disease modifying therapies did better than those who were not treated.
There are several issues with the use of historical controls. One issue is changes in the way we diagnose multiple sclerosis. In the past, MS was a clinical diagnosis. Now, we use MRI and laboratory tests to help diagnose MS. This allows for an earlier as well as more specific diagnosis of MS. Current patients diagnosed earlier with MRI may appear to do better than historical patients as the earlier diagnosis gives them more time from diagnosis to develop disability or convert to secondary progressive MS. In this case, the appearance of doing better may only be due to earlier diagnosis and not actual better outcomes. For example, compare a historical patient who reaches an EDSS of 4 in 10 years to a contemporary patient who is diagnosed five years sooner using an MRI and reaches an EDSS of 4 in 15 years. The historical patient would appear to have done worse than the contemporary patient as he reached an EDSS of 4 in 10 years rather than 15 years when in fact both patients fared the same.
More specificity in the contemporary diagnosis of MS means that historical groups of patients with MS include patients who may have other diseases which mimic MS. These diseases could have a better or worse outcome than MS. If these disorders have a worse outcome than MS (such as a disease called neuromyelitis opticus) they could make outcomes for historical patients look worse than current patients.
A second issue comparing historical patients to contemporary patients is that now we have better access to healthcare and better treatments for non – MS medical conditions. Better access to health care, including physical, occupational, and speech therapy and newer treatments for depression, infections, heart disease, and cancer should improve outcomes for contemporary compared to historical MS patients irrespective of whether they use disease modifying therapy. Could one of the reasons for better outcomes in the contemporary group in the Tedholm study (Tedeholm et al, 2012) been due to such things as better treatment of infections rather than disease modifying therapy use?
Bias is the third issue when one uses historical controls. Some types of bias affecting the use of historical controls include selection bias, exclusion bias, attrition bias, and survivorship bias. [See table 1]
Since neither historical nor contemporary patients are chosen at random from the MS population, confounding becomes the fourth issue when using historical controls. Known and unknown confounders may influence outcomes when comparing historical to contemporary patients. Unknown confounders that do not affect contemporary patients the same way as historical patients cannot be controlled for and could make it spuriously appear that contemporary patients do better than historical patients.
These issues complicate and may invalidate comparisons between how patients on disease modifying therapy do now compared to how patients did before disease modifying therapy was available. Aside from the use of historical controls, there are at least a few other known confounders which could have affected the outcome of the Swedish study (Tedeholm et al, 2012). One confounder was the non-randomized choice of treatment or no treatment and what treatment was given. A second source for confounding in this study was the different MS centers from which the comparator groups were selected.
More recent registry studies using contemporary patients remain with problems.
Brown et al. (Brown, Coles, Horakova, & et al., 2019) is a recent study which examined the association between the use of disease modifying therapy, the type of disease modifying therapy, and the timing of starting a disease modifying therapy with the risk of conversion to secondary progressive MS.
This study included 1555 treated patients identified from the MSbase Registry, 53 alemtuzumab treated patients from 5 European non-MSbase centers before it was more universally licensed in Europe, and 230 untreated patients from a tertiary referral center at the University Hospital of Wales in Southeast Wales. The MSbase Registry included 68 neurology centers in 21 countries at the time of the study. Patients from the MSbase registry were initially seen between 1988-2012 and followed, for a minimum of 4 years until 2017.
The study used a technique called propensity matching to control for some of the known likely confounders. Each treated patient was statistically matched to similar untreated MS patients. Treated patients were matched to untreated patients based on age, sex, annualized relapse rate in the year before the baseline examination, and Kurtzke Expanded Disability Status Scale (EDSS). Patients were treated on interferon beta, glatiramer acetate, fingolimod, natalizumab, or alemtuzumab. Patients on teriflunomide and dimethyl fumarate were excluded because they had less than 4 years of follow-up. When treated patients were compared to each other, they were additionally matched based on the proportion of time they were taking therapy during the first five years.
Patients treated on interferon or glatiramer acetate were restricted to those patients receiving these medications before fingolimod, natalizumab or alemtuzumab were available. This meant that patients on interferon or glatiramer acetate were from an earlier time than the other treated patients(and thus would be considered historical in nature.)
Using patients from this earlier time helped to ensure that the severity of MS in patients in the interferon/glatiramer acetate group was like that of the fingolimod, natalizumab and alemtuzumab groups. By the time fingolimod, natalizumab and alemtuzumab were available, these medications would be preferential used for more severe MS patients while interferon or glatiramer acetate would be used for patients with milder MS.
Table 3 shows the number of patients screened versus entered the study.
Table 3
Number screened | Group | Number selected |
43,048 | MSbase patients | 1555 |
1091 | Welsh untreated cohort | 230 |
78 | Alemtuzumab | 53 |
Untreated patients were compared to those on either interferons or glatiramer acetate, fingolimod, natalizumab, or alemtuzumab. Patients on interferons or glatiramer acetate were compared to those on fingolimod, natalizumab, or alemtuzumab. Patients initially started on glatiramer acetate or interferons and later switched to fingolimod, alemtuzumab, or natalizumab were compared to see if it made a difference if one switched therapy before or after five years on glatiramer acetate or interferons.
The study found: patients treated with a disease modifying therapy had a lower risk of conversion to secondary progressive MS than untreated patients (Table 4); initial treatment with fingolimod, alemtuzumab, or natalizumab was associated with a lower risk of conversion than initial treatment with glatiramer acetate or interferon beta (Table 5); the probability of conversion was lower when glatiramer acetate or interferon beta was started within 5 years of disease onset versus after 5 years of onset (Table 6); and, the risk of conversion was higher when patients who were started initially on glatiramer acetate or INF were switched to fingolimod, alemtuzumab, or natalizumab after 5 years on glatiramer acetate or interferons compared to those started before 5 years (Table 7).
Table 4
Untreated | Inf/ GA | Untreated | Fingolimod | Untreated | Natalizumab | Untreated | Alemtuzumab | |
5 years | 27 | 12 | 32 | 7 | 38 | 19 | 25 | 10 |
*57 | 47 | **39 | 7 | **48 | 34 | ***41 | 21 |
*At 11 years ** At 6 years ***At 8 years
Table 5
Inf/GA | fingolimod/natalizumab or alemtuzumab | |
5 Years | 12 | 7 |
9 Years | 27 | 16 |
Table 6
Before 5 years | After 5 years | |
5 Years | 3 | 6 |
7 Years | 29 | 47 |
Table 7
Before 5 years | After 5 years | |
5 Years | 8 | 14 |
17 Years | 14 | 28 |
This study suggests a significant effect of disease modifying therapy on long-term disability in MS. Treatment appeared to delay the onset of secondary progressive MS. Treatment effect on long term disability is further supported by the fact that the effect appeared more prominent with fingolimod, alemtuzumab, or natalizumab. with fingolimod, alemtuzumab, or natalizumab than glatiramer acetate or interferons. Fingolimod, alemtuzumab, or natalizumab are generally believed to be more effective than glatiramer acetate or interferons.
Nevertheless, the study illustrates persistent problems using registry data. One problem is the use of historical controls. Untreated patients included those from 1988 to 2012. Patients on GA and IFN were patients from MSbase but treated from 1996-1998. This means that some of the differences observed between treated and untreated patients and older and newer therapies could be due to temporal trends rather than effects of the drugs themselves.
A second problem is non-randomized treatment decisions. A wide variety of physicians managed each of these patients. Patients likely received different disease modifying therapy as well as other medical management depending on the physician’ training, background and where he was from. Each patient was treated or not treated at the discretion of the treating physician. We do not know why each patient was treated. Similarly, the choice of treatment was also at the discretion of the treating physician. We do not know why a specific treatment was chosen. This means that each group of treated patients may differ in unknown ways from each other and from the untreated patients.
Selection bias is a third problem potentially affecting the outcomes. Table 3 indicates that treated and non-treated patients were highly selected as only 1555 of 43.048 patient were included. The reason for this selectivity is the inclusion criteria which included complete background data, 3 or more EDSS measurements (one before and two after treatment, lack of baseline data, and more than four year of follow-up. Patients on two commonly used therapies, teriflunomide and dimethyl fumarate, were excluded because they had less than 4 years of follow-up.
It is important that patients selected for a study are representative of a more general population of MS patient. Only then can the results of a study be generalized to outcomes in all MS patients. Are the patients selected for this study representative of all MS patients? In order to determine this, we need to look at the groups in more detail.
The untreated patents were drawn from the neuroinflammatory service database at the University Hospital of Wales. This is a tertiary referral center in Southwest Wales. Initially, in 1985, this database stated from a population-based prevalence study in the county of South Glamorgan when 441 MS patients were identified out of a population of 376,718. (Swingler & Comston, 1988) Later, the clinical data was identified from “annual or semi-annual” appointments but we are not told in this study if this group remains geographically based and thus representative of all MS patients in the county of South Glamorgan. In 1985 the area was served by 214 general practitioners and 3 neurologists but we are not told who was treating these patients when they were chosen for the study, nor why they remained untreated.
Treated patients, except for those on alemtuzumab, were drawn from the MSbase Registry. With regards to these patients, it is worth considering who was entered and excluded from the registry and which MS patients in the registry were included or excluded from the study. Patients in the MSbase registry were from 105 centers in 29 countries. The MSbase registry enters patients from tertiary MS centers and presumably physicians who may be early adopters. Inclusion in the registry selects for patients seen in tertiary care patients who may be more rapidly progressive or have more advanced MS than the general population. The early adopters may manage MS patients differently than the general neurologist which makes their management difficult to compare to the general neurologist. The sample of 1555 out of more than 43 thousand patients in the registry also makes it less likely that the treated patients are representative of MS patients treated with disease modifying therapy in general.
The patients in the Alemtuzumab group were from five European non MSbase centers using Alemtuzumab before it was licensed (Bristol, Cardiff, Swansea, Dublin, and Dresden). This group (and the doctors who treated them before the drug was approved for general use) are clearly not representative of the universe of all MS patients and not truly comparable to the group of patients from the MSbase nor Wales untreated group. We do not know why these patients were chosen for treatment. From our experience with disease modifying therapy, we assume that they had aggressive disease but do not know this for sure. Also, these patients as well as the neurologists who treated them fall into the category of early adopters unlike most patients and neurologists.
Major confounders which are known to affect time to secondary progressive MS such as age and EDSS were matched between groups. A fourth problem with this study is that other known confounders, including socioeconomic status, were not controlled. The technique of propensity score adjustments used in this study helped to decrease the effects of some of the known confounders but cannot remove the effects of treatment decisions, different times the groups were formed, or studies are done, and geographical locations (Naismith, 2020).
Without random selection, unknown confounders also could not be controlled.
To give Brown et al. credit, although the patients chosen for this study may not be truly representative of all patients with MS, the patients chosen for this study and the techniques used are probably the best we have at this time to help answer our question of the long-term benefit of disease modifying therapy
Summary
In this post, I have reviewed clinical outcome measures used for relapsing-remitting MS and discussed some of the reasons it is difficult to prove that disease modifying therapy affects long term MS outcome. These difficulties are mainly related to the inability to do long term randomized clinical trials. This leaves us with the use of observational studies. Some examples of observational studies were presented.
Observational studies are limited due to bias and confounding which is difficult to control even using multiple regression analysis.
Nevertheless, the evidence for long-term effectiveness and safety of disease modifying therapy is building. Given the risk of serious long-term disability from active relapsing-remitting MS, the building evidence endorses the continued use of disease modifying therapy to help prevent long term disability from active relapsing-remitting MS.