Clinical trials vs. real-world data in MS


We probably don’t really require a study to tell us that those participating in clinical trials are not exactly the same as the general population that may take a particular treatment – we already know this. However, when it comes to MS drugs it is worth noting what the differences between the two populations are.

According to Rojas et al., upon reviewing 18 clinical trials and 73 real-world practice data, the clinical trial group were a cleaner population (or treatment naive) and less likely to have been on previous MS drugs. The potential for interactions is therefore substantially less in the clinical trial population.

If we think of what’s happening in the body as a chemical reaction, the chemical reactions in the clinical trial population are more likely to be starting with the base ingredients, whilst those in the real-world scenario are entering something new into the reaction half-way through. The differences add up with a greater number of permutations over the course of several years in the real-world population.

Which then begs the question, which piece of evidence is more important when considering treatment selection – the clinical trial data or the real-world evidence?

I take a two-pronged approach to this – 1) a clinical trial is definitely the area to test treatment/drug efficacy. A pure setting is needed with as little variability in the test as possible in order to answer this question; 2) following on from here, real-world evidence is where you examine the endurance/rigor of the treatment, including its long-term safety in a larger population that wouldn’t have been typically included.

How you apply clinical trial data into real-world practice should follow something similar to this approach. It is important to compare like-for-like, rather than dissimilar comparators. So the next time you come across some real-world evidence, don’t gloss over it.


Mult Scler Relat Disord. 2020 Jan 3;39:101931. doi: 10.1016/j.msard.2020.101931. [Epub ahead of print]

Do clinical trials for new disease modifying treatments include real world patients with multiple sclerosis?

Rojas JI, Pappolla A, Patrucco L, Cristiano E, Sánchez F.

We often see that clinical and demographic characteristics of real-world studies (RWS) do not differ from patients included in randomized controlled trials (RCT).


to compare clinical and demographic aspects of patients included in RCT and RWS that evaluated new disease modifying treatment in multiple sclerosis (MS).


a systematic non-language-restricted literature search of RCT and RWS that evaluated new disease modifying treatments (natalizumab, alemtuzumab, ocrelizumab, fingolimod, teriflunomide, dimethyl fumarate and cladribine) from January 2005 to January 2019. Demographic and clinical data were extracted, described and compared.


18 RCT and 73 RWS were included. We found no differences in clinical and demographic aspects between RCT and RWS except in the frequency of naïve patients included in RCT vs. RWS 65.6% (95%CI 52-74) vs. 36.4% (95%CI 21-46), respectively, (p = 0.013) at study entry, as well as for the inclusion of patients that used previous treatment 34.4% (95%CI 22-41) vs. 63.6% (95%CI 53-74) in RCT and RWS, respectively,(p = 0.007) at study entry.


We did not observe significant differences in most clinical and demographic aspects of included patients in RCT and RWS. Studies that include the full spectrum of MS patients followed in clinical practice are needed.

About the author

Neuro Doc Gnanapavan


  • Interesting. What’s your opinion towards ‘early access’ and compassionate use programs? Clearly long term safety data isn’t available yet next to potential interaction with previous DMT’s.

    • Yes you’re right, but at some point that cohort would end up being included in the real-world evidence as they would be part of treatment databases of MS units. Until then it’s a bit in limbo.

  • Theres a new thought in speeding up clinical trials by merging phase 1 and 2. To two phases. I think this has been approved in Europe moving forward. Problem with extrapolating trends from real world data is the variability of control factors that can effect the efficacy of treatment. Which is minimised in clinical trials. However as a AI specialist in machine learning whether supervised or unsupervised training models. This is no longer a issue. So you can compare apples and pears. You just need to use the latest tools.

    • Yes, but Phase III is the efficacy mark – how large does this need to be for one model to be accurate? Is your model the same as what someone else would create? How do you control for unknown biases in your model?

      The real-world evidence shouldn’t be extrapolated back to efficacy (would be inaccurate), but is more about treatment sustainability, and potential complications of multiple treatments etc This is invaluable for the clinician. And with a bit of imagination, you can see the parallels between yours and the real-world data – hopefully!



Recent Posts

Recent Comments