>> Holzgreve: The term “real-world data” is polarizing. For some it is unnecessary nonsense, for others it is the solution to many problems. Now the question arises: Where does reality lie? Who will prevail and what can we do with this data? And above all, with which approach: with general data collection à la Big Data - everything is collected and then we see what is possible with it. Or shouldn't you think carefully beforehand about what goals you have and what data and in what quality are needed for this, especially when collecting data? Or would it even be better to demand that all newly approved drugs collect data in real care in order to be able to answer the majority of the questions that will certainly arise in five or even ten years?
Behring: I am not of the opinion that every drug needs to be accompanied by data collection, ABDE. The reason for this is that it does not make sense for all medicines to collect further data after the market launch if the relevant data are already available through approval in very good RCTs that leave little doubt as to what a medicine does or does not do. The new tool for data collection accompanying the application, which was given to the G-BA, also has a certain sharpness. It will be important to define in advance exactly which evidence gaps there may be. Of course, it would be ideal if you could answer all of these questions before admission.
Holzgreve: Mr. Müller, which “real-world data” are actually relevant for you? You can make the most insane correlations with all sorts of data, starting with whether mortality increases when surgical staff wear blue or red surgical scrubs.
Müller: In the future, when we have an electronic patient file in Germany, it would actually be an instrument to be able to record and answer questions about mortality and side effects, for example, and also questions about benefit assessment. That's why one focus of our work in the ministry is to first create the conditions for this. It's certainly true that if you take a closer look, the data quality is often not that great. We see this in the medical tumor registries, where we have a lot of trouble collecting the basic data. It is not enough to simply record that a patient has died, but you have to know at what stage and under what therapies. We're actually not that far along in Germany yet. This is precisely why we at the ABDE have limited its application and want the Federal Committee to handle it responsibly. From the ministry's point of view, it would be a good idea if we restricted the doctor's freedom of treatment because he is no longer allowed to prescribe a new medication if he does not take part in the data collection. We believe this is justified and have pushed this through the departments, also because we believe it is a very sharp instrument. The same applies to the point of randomization or not. I understand that it looks a little different from the perspective of the methodologists, because of course the comparison in the randomization is qualitatively better. But in the ministry we have to think practically about how such an instrument can be used and we also have to look at how it is proportionate. From a political perspective, it cannot be the case that when a new drug becomes available, a patient is asked whether he wants to be included in an RCT and, depending on how the dice falls, he is then in the group with the new drug or not.
Holzgreve: That really spoke to me because I'm already in the older semester. Mr. Wascher, in your presentation you mentioned the old category of application observation, which used to be a good additional income. But we all actually knew that this was a smokescreen for sales and were actually happy that the AWBs had finally been abolished. They have also indicated that they actually believe it is impossible for real-world data to change the results of randomized trials if they have been properly conducted.
diaperer: My hope doesn't die anyway that the G-BA will do something different and meaningful than before under the term application observations. Using the term here was certainly not very conducive to this goal. I think that the industry knows exactly what application observations are, what they do and what they are supposedly not allowed to do. If the G-BA thinks it is bound to this terminology, then I don't see much perspective. But perhaps together we can come up with good new ideas when it comes to comparative application observations. As a second point, a quick comment about real-world: This is one of those buzzwords that doesn't explain anything at all. This is a buzzword that has about ten different definitions. The crucial point is that the term real-world practically means that admissions studies were not created in the real world. The patient representatives would have to rise to the barricades and say: What are you actually doing, expecting us to approve laboratory studies that are supposed to be used in our care? That's a term that doesn't help you at all. Now about the randomized studies: I did not say and will not say that it is impossible to obtain additional complementary findings using real-world data for randomized studies. If I already have a study, then what is added with real-world can only ever be an addition to what I have. Real-world data means data from real care, which says nothing at all about the study design. My crucial requirement for this data as part of the benefit statement is a meaningful, fair comparison. There are statistical methods that can approximate this to randomized studies. Every methodologist and theorist knows that we will never achieve it, but in practice that may not be so bad. If you want to work with propensity scores, adjustment and other things, that means the requirement for the data quality of this so-called real-world data, which you can completely forget about with routine data and for which you would otherwise, and that's why I advise the industry It is emphasized that a great deal of effort must be made.
Stegmaier: What level of evidence or what quality of evidence is sufficient? How will IQWiG react, which has already rejected so many RCT studies in the past because they did not meet the minimum quality? How will we then deal with real-world data?
diaperer: It will depend on whether we have meaningful comparisons. I am saying explicitly here that these meaningful comparisons do not always have to come from RCTs, but they must be meaningful comparisons. The question of what meaningful comparisons are means what requirements are placed on the adjustment of disruptive variants. This in turn depends on the size of the effects. If you have very large effects, you hardly need to worry about adjusting them from the start. You can simply say, it's fine. If you have medium effects, then you will have to take care of it more carefully. And if you have small effects, you will have to take care of them very carefully. I'll give you a personal statement: I can't recommend anyone not to do randomized trials for small effects. It is likely that the legislative situation and the G-BA's decision will possibly and sensibly concentrate on situations where small effects are not at stake. Because other designs might also make sense.
Behring: A comment on that. It is not that randomized controlled trials are not of good quality. You simply haven't answered the question we want. It is extremely important that you think carefully about the question for which you want to conduct this or that study. And that you then choose the right data source for the right question. The point is not that the things were rejected because the data quality in the RCT was not up to par, but rather because they somehow missed the objective.
Holzgreve: It is an actual phenomenon that patients in patient reporting outcome studies realize at some point that it is better if they say, "I'm fine," because then no one will come and bother them. This means that you have to be extremely careful whether the data collected is valid.
Griesinger: How do you imagine the simultaneous randomization of parallel or comparison groups? Mr. Müller just said that this is actually difficult to imagine with medications that have already been approved. If we take a drug that has a 90 percent response rate, you will not be able to demand simultaneous randomization in the real world or in registries. Or is there perhaps an option if there is good data quality in a cohort study in which confounders have been defined and in which historical, well-clinically described patient populations can also be identified as corresponding comparison groups.
diaperer: The crucial point is a meaningful comparison. If you are in a situation where, in connection with approvals and with AMNOG, on the basis of randomized studies or on the basis of completely different data, you come to the conclusion that it is a convincing drug with a considerable additional benefit, the question arises further studies no longer exist. The moment one can make a reliable assessment of a drug or an intervention, the question of randomized trials no longer arises. By the way, neither is the question about other studies. The situation is different, however, in which one is uncertain about hopefully relevant questions regarding the relationship between the new therapy and the previous standard. Randomizing in this situation is not a problem in principle, but it is a problem in practice. That's why I say again: the fair comparison is the decisive factor. In extreme cases and in special cases this can also be historical. Historical in the sense of perhaps not 1920, but before the drug was introduced, but please: a fair comparison. In my opinion this is quite possible, but it depends on the entire situation. The crucial point – which I criticize and which, in my opinion, is an anachronism – is to prohibit randomization at this point.
Holzgreve: If the time factor is important because you have to make a decision, would you agree that you would then use all available data, even if the evidence is not that great or if you don't even know how the data was collected?
Müller: That is the discussion about what politicians expect from the Federal Committee and what is relevant in the internal discussion in the Federal Committee. We just want the Federal Committee to receive this signal with the effort that it puts in as the leading self-administration organization with all experts, patient representatives and societies: What can patients expect from a new drug? If you then add that there is a low level of reliability because, for example, there is a cohort study and not an RCT, then that is exactly what the law says. Only if there is a noticeable trend towards 100 percent “unquantifiable” results for orphans and other drugs, then we will, on the other hand, get completely different assessments in the recommendations of the specialist societies and in the discussions with the specialist groups. We in the ministry cannot decide who is right, but this discrepancy is problematic for the legislature in the long term, I'm just pointing that out. We cannot now retreat into the position and say that we – by which I actually mean the IQWiG and the Federal Committee – want the RCT. The approval has taken a different development, you have to accept that somewhere. I understand that people cling to the RCT as the gold standard, and it will continue to be the gold standard in many indications, but if you understand a mechanism in an indication or have an intervention such as a genetic intervention, then we need other tools. Then patients also expect to be given a statement about this. That's what we're trying to promote. But I believe that real-world data, for example, will actually allow for comparisons in the future. These are simply biometric advancements.
diaperer: The 100 percent is of course only correct for very specific orphan drug groups, certainly not for the normal assessments. I think the point that irritates me a bit is that you, Mr. Müller, described very comprehensibly at the beginning of your lecture how difficult the situation is due to the increasingly advanced approvals with increasingly immature data. What should we do? The consequence is that nothing concrete can be said about these medicines. And that is shown in the drug benefit regulation with “not quantifiable”: We simply don’t know. In this respect, there is a very close connection between the increasingly immature data, which means that the statements of the IQWiG and the statements of the G-BA have to become more and more imprecise and more and more shrugging of the shoulders - if I may say that a bit flippantly. We try very hard not to write “unquantifiable” on it if possible. But if the data isn't there, the data isn't there.
Müller: This is a difficult situation, but also one that is based on the consensus of the international approval authorities. It is a situation where we have a decision at European level that a new drug should reach patients. This discussion about “patient access” is limited by pricing problems in many countries, but there is no controversial discussion on the medical side at all about the fact that we need this acceleration in many indications. We also do not see that this will lead to an increased withdrawal and revocation of approvals. Of course, there are examples where we have to make corrections because we have limited data. But I wouldn't confirm that this is a complete aberration based on the number of admissions. In the HTA area, we need to think about how we deal with this in order to continue to fulfill the goal of making price negotiations for legislators and patients about new drugs. That's the challenge now. This also means that you have to say goodbye to concepts that were set as standards, like the RCT eleven years ago. In some areas, there are biometric approaches to approaching validity across large amounts of data, even if one may not achieve it.
Speaker: Orphan drugs have a special approval status because people are prepared to accept greater uncertainty. In principle, this greater uncertainty is not reflected in the G-BA and other procedures. In principle, wouldn't it make sense to introduce a level below the reference point in order to try to classify more deeply.
diaperer: What would be less of a clue than a clue? We could certainly discuss it.
Müller: We will evaluate and refocus the orphan legislation during the Council Presidency. We want orphans to bring substantial progress for patients. That is the concern. That's why what we do is not an IQWiG or
G-BA bashing, but we want a filter function. There are orphans that bring great progress for patients, but there are also those for which we have doubts. But they cost 100,000 euros or more and have side effects and the patient also has to go to the doctor every week. This differentiation gets a little lost when we always say “not quantifiable”. That is the point.
Hoffmann: We are not having a fundamental discussion, but rather an acknowledgment of the fact that we do not have any jumps in the assessment of evidence, but rather a continuous spectrum. We need to think beyond the very clear, where it is better to know a little more than nothing at all. This is something that makes me extremely happy as a methodologist and healthcare researcher, as well as the certain sobriety with which the question is approached. On the one hand, one recognizes that there are other things below the RCT threshold, but on the other hand, this means a very high responsibility that we have with it - including the industry! What we used to do with the application observations was completely irresponsible, immoral and unethical. That's why we abolished it. But today we are in a different situation in which you have to honestly admit that you are uncertain, but still make certain decisions because there are enough good arguments. But then you have to be prepared to make corrections afterwards. This is the new situation in which there are no fundamental winners or losers. But we also have to say very clearly: If the legal data protection framework remains as it is at the moment, then we will continue to not have this data or, under certain circumstances, even worse. But if we start to say that patients are not treated well here because they are not allowed to have certain data, then the trade-off between data protection and life protection is a new one. Data protection officers are then no longer allowed to answer the question on their own, but rather they are a party. The greater importance of data from the so-called real world or supply forces us to compromise on the desire to maintain absolute data protection. I recommend using the Council Presidency for this too.
Holzgreve: I think you can leave it like that. You just said that cancer registries are even regulated by law and yet we have a lot of problems with them and basically can't do anything with them. Does it even make sense to create registers on your own initiative? Are there good registries?
diaperer: There are good registers. As part of the G-BA's order, we spoke to a number of registry operators who have really good registers. But I won't mention them now because they will be part of the content of our report. The problem with good registries is less data protection than quality. You have to deliver a structural quality, in terms of completeness and such banal things. There are definitely registers that you can really do something with. But there are also a whole series of registers that you can't do much with at the moment.
Speaker: I have a quick question about the legislative process: Which patient groups or medications are affected? You mentioned orphan drugs in your talk, but then you also talked about precision medicine. Does this also mean that it will not just be the pure orphan disease group, but that genetically defined patient groups will also be included?
Müller: The text of the law is aimed at orphans and approvals for which data is missing. These are approvals such as Conditional Approval and Exceptional Circumstances. In this respect, other groups are also included. Conditionals typically refer to those cases for which data are still pending because the phase 3 studies are still ongoing. <<
Citation note:
