#TeachSpeak: our belief in small numbers

Why do we have so many cognitive frailties; e.g. our belief in small numbers? #TeachSpeak

I want to restate what I said yesterday in a comment that the mistake I made, and many others are making, when considering the cancer risk with ocrelizumab is that we are over interpreting data based on very small numbers. To understand this phenomenon I suggest you read the classic, and probably one of the most influential, papers in psychology by Tversky and Kahneman on the ‘belief in the law of small numbers’. The problem with the human brain is that we believe in small numbers. What we really need is slow thinking and large numbers, for that we will have to wait some time for more safety data to emerge regarding rare adverse events. 

One of  the reasons for keeping the survey below open so long is that I want a large sample; I need several hundred respondents for the results to have face validity and, obviously, to move away from small numbers.  

TVERSKY & KAHNEMAN. BELIEF IN THE LAW OF SMALL NUMBERS. Psychological Bulletin, 1971, Vol. 76, No. 2. 105-110. 

People have erroneous intuitions about the laws of chance. In particular, they regard a sample randomly drawn from a population as highly representative, that is, similar to the population in all essential characteristics. The prevalence of the belief and its unfortunate consequences for psvchological research are illustrated by the responses of professional psychologists to a questionnaire conceming research decisions.

CoI: multiple

About the author

Prof G

Professor of Neurology, Barts & The London. MS & Preventive Neurology thinker, blogger, runner, vegetable gardener, husband, father, cook and wine & food lover.


  • What has Psychology got to do with Ocrelizumab for PPMS? answer is nothing stop pushing into something else that has nothing at all do to with it .Sad .

    • Re: "What has Psychology got to do with Ocrelizumab for PPMS?"

      Everything; we all think, interpret and predict. Psychologists tell us how we do it and the biases we have when we do it. Scientists and neurologists are not exempt.

      A good paper to read if you are interested is this one:

      Redelmeier DA, Tversky A. Discrepancy between medical decisions for individual patients and for groups. N Engl J Med. 1990 Apr 19;322(16):1162-4.


  • That 2011 paper you use to back up your theraputic lag theory used "63 and 59 MSers"… Bit hypocritical? You cannot reassure people with sincerity about Ocrelizumab until the data is there. End of.

    • Therapeutic lag is based on quite a large, and expanding, data set from trials in the last 25+ years:

      methotrexate CPMS
      interferon-beta-1a SPMS
      interferon-beta-1b PPMS
      GA PPMS
      Siponimod SPMS
      natalizumab RRMS
      natalizumab SPMS
      rituximab PPMS
      ocrelizumab RRMS
      ocrelizumab PPMS

      Not sure where you get 63 and 59 from. Have you read the paper?

    • "MAIN OUTCOME MEASURES: After 5 years without treatment, the EDSS and MSFC measures were scored for 63 and 59 patients, respectively. Neuropsychological and magnetic resonance imaging assessments were performed for 59 and 50 patients, respectively."

    • Ah… but this study is only one example in PPMS. You need to read the literature and the LENDAXON paper. This is not a theory based on one study. The most compelling study is probably the ASCEND (Natalizumab) trial and its extension study.

    • Gavin, doesn't the therapeutic lag theory imply that something like HSCT (which eliminates all autoimmune activity), even if used on someone with SPMS or PPMS, would in 100% of cases eventually halt further disability? Obviously, the damaged existing pathways will continue to degrade but once that has burned out they won't continue to get worse.

      So shouldn't we go and look for people who had HSCT in 1995 for example and find out if any of them are still accumulating disability?

    • Good point, and would we worth a look but HCST may not eliminate immune response within the CNS, i stand to be corrected but not sure OCB go. Also how long does it take for a nerve to die back,

    • Re: "Gavin, doesn't the therapeutic lag theory imply that something like HSCT (which eliminates all autoimmune activity), even if used on someone with SPMS or PPMS, would in 100% of cases eventually halt further disability?"

      Unfortunately, not. If you acquire too much damage, and lose too much reserve, you are likely to trigger premature ageing mechanisms and hence some of the delayed slow neuroaxonal loss we see is ageing.

      When you have very little left in a particular neuronal system, ageing mechanisms that really begin at about 35 years of age are seen much sooner. This is why we push early effective treatment and time is brain so actively.

    • "Unfortunately, not."

      But then that negates your entire argument, doesn't it? If I, as a PPMSer, have lost too much reserve, HSCT will do nothing, natalizumab will do nothing, beta interferon will certainly do nothing. So we're back at square one.

      We need neuroprotectives. Early in the disease. Not more immune system hammers, not more guff.

    • Small I see you are one of the new generation of nihilists who don't think it is worth saving hand and arm function in people with more advanced MS?

      The whole point about the length-dependent axonopathy hypothesis and therapeutic lag is that even if you have lost function in the limbs you can still do something about the arms. The arms have shorter axons with more reserve and are hence relatively protected from the ravages of the disease.


    • No I am not a "nihilist". You know nothing about me, so please drop your obtuse caricatures. I have PPMS, and I just want therapies that work. I am not sure my progression is following your pretty diagrams and theories.

  • People generally are not very good at judging risk. There were far more people killed in the Grenfell tower fire than have been killed by terrorists this year – but far more people would be scared of being killed by a terrorist than by a house fire. There are plenty more examples, that's just one.

    If, by a statistical fluke, there were more people who would have got cancer anyway in the ocrelizumab arm of the trial, they will skew the results as the number of trial participants is so small that statistical anomalies will stand out.

    Breast cancer survival rates are very good if discovered early (and as a known risk, I am assuming that screening will be offered as part of the regime).

    If you take ocrelizumab, you may get cancer – but if you don', your MS will probably be worse. And if you were unfortunate enough to get cancer – was it due to the O or would you have got it anyway? On a personal level, you can never know.

    I know there is a risk, but life is risk and carries a 100% mortality rate. As long as you are aware of the risk and make an informed choice about it, what is the problem?

    No-one's going to MAKE you take O if you don't want to.

    • It is all about framing. If you frame ocrelizumab as the first effective treatment in PPMS that has twice the effect on upper limb function compared to lower limb function, it improves QoL, slows brain volume loss, delays the need for a wheelchair on average by 4 years, etc. most people will take ocrelziumab.

      If you frame ocrelizumab as a drug that depletes B cells and puts you at risk of developing breast and other cancers and is immunosuppressive and causes opportunistic infections, etc. a lot of people won't take it.

      Do you want to frame ocrelizumab as DMT or dangerous immunosuppressive? It is your call, but you also need to keep in mind how you frame PPMS.

  • "but far more people would be scared of being killed by a terrorist than by a house fire"

    That's not true.

  • Thanks for keeping this survey open. Can't see why EMA wouldn't license although not sure about NICE. Only so much £ in the pot after all and so many patient groups screaming for it.
    As to whether Roche would be interested in differential pricing? Depends if they judge it worth their while to sell x more doses at y reduced cost.

    To be honest, seems more sensible to me to spread the word to neuros everywhere to prescribe off-label cladribine for everyone with worsening or more advanced MS. Might as well play Pharma at their own game, nothing licensed for SPMS and PPMS therefore off-label generic cladribine goes. pwSPMS and pwPPMS get an effective treatment (probably the best yet, actually gets into the CNS!) and as a little bonus payers eg cash-strapped NHS save money.

    #Cladribine4MS #MSchariot

By Prof G



Recent Posts

Recent Comments