Etudes & Biais

  • RSS
  • Accueil
  • A propos
  • Lecture critique
  • Plainte vs EMA
  • Liens
  • Contact

Suite des relations avec l’EMA

Posted on 20 novembre 2013 by Alexis in Article, Relations EMA No Comments
  • Home
  • Article
  • Suite des relations avec l’EMA

Il y a un peu plus de 2 mois, j’avais reçu la réponse du médiateur européen à ma requête de document auprès de l’EMA.

Cette réponse est disponible ici.

Les points 31 et 32 de cette réponse m’ont incité à poser quelques questions supplémentaires à l’EMA.

Le texte de mon mail est disponible ici

Je viens de recevoir la réponse à ce mail. Un point m’a un peu choqué : L’EMA considère que les informations nouvelles sur l’étude Avonex ne modifient pas le rapport bénéfice/risque du produit.

La réponse de l’EMA est disponible ici.

J’ai donc posé des questions plus précises à l’EMA. Mon mail est copié ci-dessous.

L’EMA a deux mois pour répondre.

__________________

Dear…

 

Thank you for your answer.

Please find hereunder my responses to your letter attached to the mail.

I will mainly answer to the fact that the “CHMP rapporteur and co-rapporteur did not find any compelling evidence showing an impact on the benefit/risk balance of Avonex”.

One other point will then be addressed.

 

1)      “CHMP rapporteur and co-rapporteur did not find any compelling evidence showing an impact on the benefit/risk balance of Avonex”.

This sentence in your answer almost “frightens me” as it implies that the CHMP does not consider that the trial is biased and thus, that other biased trials might easily be accepted in 2013 for product approval.

Let us consider that the benefit/risk ratio can be negatively impacted if the benefit is decreased or the risk is increased. My point is only dealing with the efficacy part of the ratio. Avonex efficacy can be evaluated on relapses rate or disability progression.

 

→     Concerning the evaluation on exacerbation rates :

 

Here are the evaluations made in the SPC and in the scientific discussion for the exacerbation outcome in the Avonex efficacy trial (MSCRG).

 

  • SPC : “During the first year there was no statistically significant difference between the two groups in number of exacerbations. Over two years a statistically significant difference was seen between the two groups in exacerbation rate (p=0.002). During the second year the exacerbation rate difference between the two groups increased though there was no statistically significant difference in the proportion of patients who were exacerbation-free.”

 

  • Scientific discussion : “It was also shown that there was a one-third reduction in annual relapse rate. This latter clinical effect was observed after more than one year of treatment.”

 

Evaluating the trial, you thus clearly consider that the results on the cohort followed for two years are valid and correctly assessing the efficacy of Avonex on the relapse rate.

 

If you had asked for the missing data in 1997, you would possibly have noticed that during the second year the exacerbation rate difference between the two groups is not increasing contrary to what the sponsor said in the Clinical Study Report (CSR) and to what you state in the SPC. Patients followed for two years had the same “benefit” on the relapse rate during the first year while patients followed for less time were doing worse when treated with Avonex. All needed data are in table 2 of my article. This information is new for you as it was not given to you in the approval file.

 

 

Considering as valid the results on the subset of patients followed for two years, knowing now the data in the different cohorts, is in fact unbelievable.

It implies that :

  • EMA will accept results based on a subgroup of randomized patients obtained after exclusion of patients, for whom the verum has a deleterious effect.  Furthermore, you consider that the results on this subgroup should apply to the whole population of patients to be treated. Pharmaceutical Industry should be informed of that : they can exclude placebo patients doing well and verum patients doing bad to get a good trial and a good p value to get approval, CHMP will accept these data! Do you think it is a fair way to analyze a clinical trial?
  • EMA validate data on the best responders without evaluating who are the patients doing worse on the verum. Can we speak of any benefit for those patients? What is the benefit/risk for these patients? What has been done to evaluate if some patients should not receive Avonex?
  • Furthermore, you accepted a CSR where the sponsor is hiding information to you, when it says that the efficacy on the relapse rate is increasing (see CSR). As I discussed with you during our teleconference call, The sponsor surely checked the efficacy on relapse rate during the second year. Knowing that the benefits were -10% for all patients during the first year and -32% for patients followed for 2 years during the two years, the sponsor surely checked that the efficacy during the second year was not over -32% around -50%. This would have been a huge marketing advantage over competitors. Furthermore, in 1997, the FDA had already told the sponsor that the benefit on relapse rate had not increased and according to your letter page 3, the sponsor “has the obligation to inform the commission and the EMA of… new information which might influence the evaluation of benefit and risks”. Was it done? Do you consider that this information does not change the evaluation of the efficacy of Avonex on relapse rate? If yes, thank you for explaining your answer.

 

Thank you for confirming (or not) those points as it is an important information for the pharmaceutical industry and of course for the European patients. Furthermore, it is now a 2013 point of view, not an old problem.

 

→     Concerning the disability outcome, you also consider that the efficacy of Avonex is valid.

 

The difference between the different cohorts is obvious on the relapse rate outcome, which is a secondary outcome.

On the primary outcome, the two cohorts (patients followed for one year and patients followed up for two years) also show differences at the end of the first year : patients followed-up for two years seem to benefit from Avonex, those followed up for less than two years seem to gain no benefit from Avonex. Data are in table 4 of my article. Patients followed up for two years are thus also different on the primary outcome at the end of the first year from those followed up for one year.

The primary outcome is based on delays (time to sustained progression in disability). A longer follow-up will favor the validation of the primary outcome.

If among randomized patients, those favoring Avonex on the exacerbation rate outcome (and also on the disability outcome) are followed-up for a longer duration, it is highly probable that the bias demonstrated on the exacerbation rate outcome will also favor the primary outcome, especially in early RRMS and during a short duration trial, (http://www.ncbi.nlm.nih.gov/pubmed/21622592) which is the case in the Avonex MSCRG trial.

 

I hope that, if you had seen the difference on the relapse rate outcome, you would have considered that the bias could have also impacted the primary outcome and at least asked for an evaluation of the baseline characteristics of the two cohorts, for an evaluation of the possibility for a censored patients to validate the disability at the next scheduled visit and checked if censored patients were obtained accordingly to the recruitment schedule, just to see if it was a bias or a fraud. Now that you know the bias, how do you explain that it does not change anything for the evaluation of the efficacy of Avonex?

 

 

→     Considering the benefit/risk ratio :

 

Saying that the benefit/ratio is not impacted for Avonex in relapsing remitting multiple sclerosis after being informed of the different cohorts is then surprising.

 

As stated in your 1997 evaluation, the approved benefit is -32% on relapses rate. As this result is based on a specific cohort, It should thus also be stated that for some other patients, Avonex may have a deleterious effects. For those patients followed up for a shorter duration, Avonex increases the relapse rate and of course provide no benefit on the disability outcome. Is it ethical to hide this information for the patients ? Accepting the -32% figure is justified by an increase of the benefit of Avonex with treatment duration, which is untrue !  It is also of course untrue to state that patients should take Avonex for more than one year to obtain a benefit ! I recognize that you did not state this point specifically, but it is a consequence of what you stated.

Consequently, it is difficult to consider that you still validate a benefit on relapses of -32%.

The bias also impacts disability progression outcome which should be also re-evaluated when defining the benefit/risk ratio.

 

Could you then explain why the new information on the different cohorts does not change for you the evaluation of efficacy and what is the benefit that you consider best represents the efficacy of Avonex for both outcomes?  Why do you still think that untrue characteristics of the trial are still valid?

 

More generally speaking, truncated trials warrant deeper analysis and if I may, I would suggest to create guidelines on how to present the results of truncated trials in a CSR.

 

Please also note than the Cochrane Collaboration, with a stringent analysis of truncated trials, recently considered that the benefit risk ratio for Avonex was unfavorable : “The lack of convincing efficacy data shows that IFNß-1a (Avonex), intravenous immunoglobulins, cyclophosphamide and long-term steroids have an unfavourable benefit-risk balance in RRMS”. (http://www.ncbi.nlm.nih.gov/pubmed/23744561)

 

2)       “Comments not published by the EMA”

 

Thank you for explaining why it is stated in the model you provided for comments that : Please note that these comments and the identity of the sender will be published unless a specific justified objection is received.?

What was the justified objection you received for not publishing my comments?

 

To conclude :

 

All your answers are important. They will provide Pharmaceutical industry and European patients with a clearer view on the way EMA analyzes trials in 2013.

Unfortunately, these stakeholders, reading your answers, might not react in the same way.

 

 

Share

Leave a Reply Cancel reply

Votre adresse de messagerie ne sera pas publiée. Les champs obligatoires sont indiqués avec *

Search

Commentaires récents

  • Alexis dans Lecture critique (3) : Analyse de la population incluse et ajustements.
  • FRG dans Lecture critique (3) : Analyse de la population incluse et ajustements.
  • Daniel Pérez dans La vérole et le bar-tabac
  • Azaïs dans Je viens de télécharger les analyses de la FDA sur Tecfidera ®
  • Alexis dans Chic ! Avec l’intelligence artificielle, on va avoir plein d’études biaisées.

Articles récents

  • Chic ! Avec l’intelligence artificielle, on va avoir plein d’études biaisées.

    3 avril 2018
  • Deux ans sans publication mais pour la bonne cause

    13 mars 2018
  • Quand il y a un « trou » (dans les données), c’est qu’y a un loup ! (Martine et Alexis)

    19 juin 2015
  • J’adore supputer, ce sera donc un peu moins sérieux… quoi que…

    21 mai 2015
  • Deux autres petits trucs qui rendent les odds-ratios un peu « odd ».

    14 mai 2015
  • Bon la dompéridone on n’y croit pas trop…et qu’est ce qui la différencie du mediator ?

    1 mai 2015
  • La vérole et le bar-tabac

    19 avril 2015
  • Domperidone

    14 avril 2015
  • Propensity scores, ou comment transformer une étude de suivi en un essai randomisé.

    3 avril 2015
  • The Will Rogers phenomenon, et petite digression sur le dépistage

    9 février 2015
  • Tiens, la National Academy Press publie un document sur les données des essais cliniques

    5 février 2015
  • Le monitoring nouveau des essais cliniques est sorti. Bonne ou mauvaise nouvelle pour les fraudeurs (et pour ceux qui les recherchent) ?

    30 janvier 2015
  • Faire une étude sur cohorte positive, fastoche !

    27 janvier 2015
  • Fraude et essai clinique biaisé. La détection des fraudes permet-elle d’exclure les essais cliniques biaisés ?

    6 janvier 2015
  • Les méta-analyses ne seraient-elles pas réellement intéressantes quand elles sont non-valides en raison d’une hétérogénéité ?

    5 décembre 2014
  • Lecture critique (12) Analyse des essais cliniques randomisés en double aveugle destinés à mettre en évidence une supériorité: récapitulatif

    20 novembre 2014
  • Finalement, il ne s’est rien passé à l’EMA ce jeudi 2 octobre.

    7 octobre 2014
  • Si on commence à critiquer les avis de l’EMA …

    28 février 2014
  • Les courbes de survie

    18 février 2014
  • Réponse de l’EMA à mon mail du 20 Novembre 2013 … et ma réponse

    9 février 2014
  • 20 novembre 2013 – 21 janvier 2014 : nouvelle plainte contre l’EMA

    21 janvier 2014
  • La FDA, ses analyses sélectives et l’Alemtuzumab

    6 janvier 2014
  • NNT

    18 décembre 2013
  • Incidence et Prévalence

    13 décembre 2013
  • Risque de sclérose en plaques avec les vaccins (ou autre produit)

    27 novembre 2013
  • Suite des relations avec l’EMA

    20 novembre 2013
  • A la recherche du bon critère d’évaluation

    18 octobre 2013
  • Evénements indésirables rares et loi de Poisson (ou loi binomiale)

    11 octobre 2013
  • Policy 0070 on publication and access to clinical-trial data

    3 septembre 2013
  • Lecture critique (11) : biais ou risque de biais,

    30 août 2013
  • Ô temps, suspends ton vol ! … de données valides

    16 juillet 2013
  • Individual Patient Data (IPD) : une nécessaire mise à disposition de la communité scientifique

    11 juillet 2013
  • P…. d’effet placebo !

    2 juillet 2013
  • Règles CONSORT pour les essais tronqués

    27 juin 2013
  • Lecture critique (10) : Les analyses de sous-groupes

    26 juin 2013
  • Alltrials pour les biais de publication… et les autres biais?

    19 juin 2013
  • Lecture critique (9) : L’aveugle dans les essais cliniques.

    10 juin 2013
  • Analyse critique (8) : La cohérence externe

    5 juin 2013
  • Je viens de télécharger les analyses de la FDA sur Tecfidera ®

    18 mai 2013
  • La puissance

    12 mai 2013

(c) 2013 Etudes & Biais -