COVID-19 and CD20: The Wizard of Odds!


Our Italian colleagues (definitely not wearing the kind of suit depicted above) were able to gather the troops, and publish their data on COVID-19 severity and the use of different disease-modifying treatments (DMTs) in people with multiple sclerosis (pwMS).

They collected data on 844 pwMS on variable DMTs. After putting all the potential risk factors for a worse COVID-19 disease course in one statistical model, they concluded there were three risk factors associated with a severe covid-19 disease course: 

  1. Age (unfortunately risk factor for almost everything bad in medicine) – Odds Ratio 1.06
  2. Usage of methylprednisolone in the preceding month – Odds Ratio 5.24
  3. Treatment with anti-CD20 (read: ocrelizumab or rituximab) – Odds Ratio 2.37

How do we interpret these results?

The Odds Ratio (OR) represents the odds that an outcome (=COVID-19 severity) will occur given a particular exposure (=DMT or anti-CD20) , compared to the odds of the outcome occurring in the absence of that exposure (= No DMT).

An OR of 2.37 for CD20 therapies means that your risk of ending up with medium/high COVID-19 severity when being on anti-CD20 is approximately double compared to untreated pwMS. 

Absolute risks are difficult to interpret, and do not allow to disentangle which part of the increased risk is conveyed by anti-CD20 and which part by other variables such as age, disability etc. To avoid this pitfall (partially), the researchers have only compared absolute risks in pwMS that were < 65 years old and EDSS < 6.5. They also made a distinction between progressive and relapsing phenotypes:

The odds behind the odds: 

  • The biggest problem to interpret these results correctly is the fact that is is not a population-based study. PwMS were included in this study if their neurologist became aware of their COVID-19 disease. You can already imagine that this way of recruiting is intrinsically biased towards pwMS on higher-efficacy therapies. These pwMS are generally more anxious to have a severe COVID-19 disease course and are much more likely to report symptoms. Nonetheless, this also accounts for people on natalizumab.
  • This type of “observational case-series” leads to important differences between DMT groups. Ocrelizumab is the only drug licensed to date for pwMS in the progressive disease stages. Untreated pwMS tend to be older, more progressive and disabled than pwMS on DMTs. People on other DMTs than anti-CD20 tend to be younger, relapsing and less disabled. The problem is that the OR represent a comparison between these unmatched groups. The researchers try to compensate for these differences in baseline characteristics by using a propensity-matched analysis. However, this is a statistical approximation and does not necessarily represent the real-life situation. 
  • Although the total size of the cohort is substantial, there was a low incidence of the hard outcomes. In total, “only” 13 pwMS died and apart from two individuals they were all progressive MS patients in an advanced disease stage. Of these 13 deceased pwMS, there were 2 pwMS treated with anti-CD20. Moreover, “only” 38/844 pwMS ended up in the ICU (unclear how many anti-CD20 in this subgroup). Importantly, low absolute numbers of events per DMT group can give rist to big jumps in the absolute risk percentages. When the statistics were repeated to assess risks associated with the hard mortality/ICU outcomes, no statistical significance could be reached. This implies that 1) there is no effect on mortality OR 2) the effect on mortality is too small to be detected by this study population (most likely). 


Although there is a signal with COVID-19 severity and anti-CD20 therapy, this study most likely represents an overestimation of the risk. The results of a 2019 population-based study in Sweden are probably more indicative of the real infection risk associated with anti-CD20. In 6421 pwMS, Luna et al. reported an hazard ratio of 1.70 of Rituximab compared to injectables. Overall, I feel that the potential increase and – worst-case scenario – doubling of the risk of having a more severe COVID-19 disease course is not the most important issue when it comes to anti-CD20 treatment. The real predicament is the reduced vaccination response and the issues around timing of vaccines and scheduling the 6-monthly treatment cycles. My bet is that COVID-19 will stay around for the remainder of our days, and – in line with the flu – will require annual repeat vaccinations. How this is compatible with a therapy that consistently blunts vaccine responses in a significant proportion of treated patients is unclear to me. TBC. 

Disease modifying therapies and Covid-19 severity in Multiple Sclerosis

Maria Pia Sormani, Nicola De Rossi, Irene Schiavetti, Luca Carmisciano, Cinzia Cordioli, Lucia Moiola, Marta Radaelli, Paolo Immovilli, Marco Capobianco, Maria Trojano, Paola Zaratin, Gioacchino Tedeschi, Giancarlo Comi, Mario Alberto Battaglia, Francesco Patti, Marco Salvetti, Musc-19 study group

PMID: 33480077 DOI: 10.1002/ana.26028


Objective: To assess the impact of immunosuppressive and immunomodulatory therapies on the severity of Coronavirus disease 2019 (Covid-19) in people with MS (PwMS).

Methods: We retrospectively collected data of PwMS with suspected or confirmed Covid-19. All the patients had complete follow up to death or recovery. Severe Covid-19 was defined by a 3-level variable: mild disease not requiring hospitalization vs pneumonia or hospitalization vs Intensive Care Unit (ICU) admission or death. We evaluated baseline characteristics and MS therapies associated to a severe Covid-19 by multivariable and Propensity-Score (PS)-weighted ordinal logistic models. Sensitivity analyses were run to confirm the results.

Results: Of 844 PwMS with suspected (n = 565) or confirmed (n = 279) Covid-19, 13 (1.54%) died: 11 of them were in a progressive MS phase, and 8 were without any therapy. Thirty-eight (4.5%) were admitted to an ICU; 99 (11.7%) had a radiologically documented pneumonia; 96 (11.4%) were hospitalized. After adjusting for region, age, sex, progressive MS course, EDSS, disease duration, BMI, comorbidities and recent methylprednisolone use, the therapy with an anti-CD20 agent (Ocrelizumab or Rituximab) was significantly associated (OR = 2.37,95%CI = 1.18-4.74,p = 0.015) with an increased risk of severe Covid-19. Recent use (<1 month) of methylprednisolone was also associated with a worse outcome (OR = 5.24,95%CI = 2.20-12.53, p = 0.001). Results were confirmed by the PS-weighted and by all the sensitivity analyses.

Interpretation: This study showed an acceptable level of safety of therapies with a broad array of mechanisms of action. However, some specific elements of risk emerged. These will need to be considered while the Covid-19 pandemic is persisting. This article is protected by copyright. All rights reserved.

About the author

Ide Smets


  • Thanks a lot.

    I would be cautious interpreting the absolute risks in particular – taken at face value we anti CD20 patients should never go outside. Very likely we are missing a good number of very mild cases that either were not diagnosed at all or never reported to the neuro. How many, hard to say.

    My going assumption remains that despite ocrelizumab, my risk is 2-3 times that of the age matched base population. Which is bearable given adequate mitigation measures and that I am below 40.

    • Totally agree, absolute risks are treacherous and the percentages shown in the graph are daunting. The study design is indeed prone to missing mild cases, but as there was no signal with natalizumab it cannot explain everything. Age and disability are definitely the most important factors.

  • I am wondering why the OR for age is so low, 1.06. How can this be given that age is supposed to be the most important risk factor, even if only <65 are included.
    Also anti-cd20 increases the risks for other infections. Are other infections not overrepresented in the non-confirmed cases for anti-cd20?

    • I also thought about this, but I think it’s because the effect of disability in this pwMS cohort probably exceeds that of age. PwMS have higher disability levels at a given age than peers without MS. The OR is calculated based on the full cohort, which also includes pwMS > 65 yr. It’s possible that the suspected cases represent other infections but this also accounts for all the other treatments, and the authors did a separate analysis specially in the confirmed cases which confirmed the results of the full cohort.

      • I can explain this. The OR depends on the unit of measure. For age in this model it is 1 year. So the OR=1.06 mean a 6% increase in risk per year of age. With a 10-year unit of measure the OR would have beeen 1,8; with a 20 year unit of measure it would have been 3,3. This indicates that the risk becomes 3 times higher every 20 years of additional age

By Ide Smets



Recent Posts

Recent Comments