Dr. Thomas Müller, head of the Medicines Department of the Federal Joint Committee, studied pharmacy and medicine in Berlin and was most recently director of the central pharmacy and head of medical procurement at the Rostock AöR University Hospital, also responsible for the purchasing of medicines. Müller, who is responsible for the internal preparation of all drug assessments at the Federal Committee, knows Hans-Holger Bleß, who is actually the head of the health services research department at the IGES Institute, but is actually directly or indirectly involved with a large number of drug dossiers, very well: The two meet in their work regularly with each other, but with very different views on what is happening at AMNOG, as becomes quite clear in the double interview with “Monitor Healthcare Research”.

Around two and a half years ago, IQWiG started a two-arm pilot project: two teams were supposed to analyze the possible uses of two economic models - on the one hand the Analytic Hierarchy Process (AHP), on the other hand the Conjoint Analysis (CA) - in two different indications. About a year ago, IQWiG published the working paper on the AHP by Prof. Jizerman, University of Twente, and a few weeks ago that by Prof. Mühlbacher, Neubrandenburg University of Applied Sciences, on the CA. When the CA paper was published, IQWiG wrote: “The method of conjoint analysis is in principle suitable for finding out what preferences patients have regarding therapy goals.” Mr. Müller, under what premises can you imagine that the G-BA recognizes these two or one of the two models in order to be able to take patient preferences into account in the AMNOG process more than before?
Müller: It is important to know that this is a methodological proposal that comes from the area of cost-benefit assessment, in which the QALY and the efficiency frontier are the two main approaches. Both approaches require an aggregated benefit measure and, in turn, a methodology that brings together the various endpoints and effects of active ingredients and therapies. Conjoint Analysis and the Analytic Hierarchy Process are survey approaches from whose results an overall measure is formed using statistical methods.

Bless: In addition, it should be mentioned that the G-BA has not yet carried out the cost-benefit assessment once in the AMNOG process.

Müller: The legislature has placed the cost-benefit assessment at the end of the AMNOG process. It can be carried out, but it is not yet because neither a pharmaceutical company nor the GKV-SV have applied for one.

What do you think of the methodology itself?
Müller: I am rather skeptical about these statistical methods because any patient preference is always an individual value decision. We should be careful not to use statistical methods to convert individual illness experiences into a kind of majority decision - which in my view is not appropriate.

Bless: Nevertheless, such value decisions are inherent to the process, since results with different directions of effect and in different categories such as survival, morbidity, quality of life and undesirable effects must be balanced to form an overall result.

Müller: In the G-BA, no grading evaluations are made or hierarchies of end points are created.

Would you give an example?
Müller: When preparing the decision, we place end points next to each other, as required by the legislator in the Drug Benefits Ordinance. For example, it may well be that for one disease, a month's longer survival is crucial for the patient, while for another, it is crucial to avoid or relieve itching. These cannot be offset against each other.

Conversely, this means that the G-BA is not concerned with the question of whether survival is more important than quality of life in an illness.
Müller: Only in the event that we recognize that an active ingredient has positive effects on one endpoint and negative, i.e. opposite, effects on another, does the G-BA check whether this also needs to be reflected in the overall result. The G-BA then takes this into account by balancing. The overall positive and negative effects are considered and finally evaluated in the overall result.

This netting sounds like an exact science. But in reality you will probably “balance” apples and oranges.
Müller: It is precisely for this reason that there cannot be an algorithm, neither in the sense of a calculation methodology nor an empirical survey, which IQWIG tested with these two model experiments. For the G-BA, this procedure is and remains an individual value decision that must be made on the basis of evidence and statements.

Bless: This individual value decision is made by a committee of people who are not directly affected by the disease and do not represent the patient's perspective. In return, at least the sending institutions are affected by the consequences of their decisions.

Let's understand this: On the one hand, the IQWiG and the G-BA maintain the highest standard of evidence in the form of RCT studies and on the other hand, eminence-based individual decisions are made, so to speak. That's quite a big spread.
Müller: The G-BA plenary session is tasked with evaluating new active ingredients at an abstract population level in order to prepare price negotiations on an evidence basis. This must be distinguished from an individual decision at the doctor-patient level - this is not restricted in any way by the G-BA. Each patient thinks individually with their doctor about which therapy is suitable for them in their very specific situation.

Bless: In doing so, the patient and the doctor are simply making a value decision for themselves.

Müller: Exactly. That's why the G-BA wants to provide doctors and patients with the best possible information relevant to decision-making about all expected effects. With oncological therapy, there may be a chance that a patient can live seven months longer with a special medication, for example, but this also comes with certain side effects. This balancing decision must be made again and again by the doctor and patient in every such therapeutic situation.

Bless: That would be the ideal. But you never come to a group decision that way.

Müller: The Federal Committee is faced with the challenge of evaluating products in order to define prices based on these decisions. Finished medicines in the GKV only have one price for all patients. Here we certainly see the difficulty of aggregating the various effects in order to arrive at a uniform assessment result and, based on this, a uniform reimbursement amount. However, this evaluation result in no way replaces the individual therapy decision.

Bless: But that is exactly the approach of IQWiG with conjoint analysis or AHP. This is precisely where the previously unused options for measuring preferences lie.

Müller: And the NICE approach is QALY. In Germany we do not yet have an algorithm based on the NICE model with an ICER and a QALY calculation. What we do here is a kind of counter-calculation: if the evidence is significant for both harm and benefit, it is balanced.

Now you suddenly calculate against what you just said, there is no algorithm.
Müller: This is due to the categories of the benefit regulation. The legislature has specified three positive categories, with “cure” being mentioned in the highest category as “significant additional benefit”. There is no scale from 1 to 100, but rather a semi-quantitative approach. Qualitative characteristics such as “sustainable”, “particularly pronounced”, “moderate” or “low” effects are netted. Overall, there is a very firmly established position in the G-BA that empirical surveys cannot replace the value decisions made in this way. I am also very skeptical as to whether the possible range of individual value decisions can be processed statistically in the same way as scientific evidence.

Bless: The G-BA makes normative decisions that replicate on a cohort that in principle covers the whole of Germany. That's why the Federal Committee would ultimately have to use appropriate instruments that could statistically record this majority. Making purely individual decisions is a bit too little for me here.

Müller: I have to clearly disagree here. These individual value decisions cannot follow a majority. The G-BA cannot make a value decision for the individual patient.

But the G-BA does this paternalistically for all patients. Wouldn't it be better to at least try to reach a majority decision using conjoint analysis or another statistical method?
Müller: This line of research is certainly important in order to make informed decisions. I think empirical research into how patients evaluate illnesses is essential. This approach is particularly valuable as a valid basis for patient representation in the Federal Committee. Nevertheless: If you use statistical methods to find out that, for example, 70 percent of those surveyed are of the opinion that this or that end point is particularly important, that cannot at the same time mean that the other end points required by the legislature are to be neglected. Because for 30 percent of patients, exactly the other endpoints can be valuable. This is not a decision that the G-BA makes paternalistically.

Mr Bleß, you winced at the word “paternalistic”.
Bless: Yes, because the patient relevance, which is so important in the process, is decided on a paternalistic basis instead of involving and questioning the patient. We have been creating a whole series of such preference measurements at the IGES Institute for years, for which we mostly use discrete choice analysis, which in turn is a variant of conjoint analysis. In principle, we do exactly what the G-BA does: We measure balancing and balancing decisions, but from the patient's perspective. This records which side effects are accepted in order to achieve which positive effect and of course negative and positive effects are taken into account. This preference measurement answers exactly the question that Mr. Müller asked and that the G-BA constantly has to ask itself.

Which one?
Bless: There are always different aspects of benefit when evaluating a new drug, for example a minor effect on quality of life, mortality or morbidity. So the question is: What becomes of three times minor? Does it become significant or significant?

The completely honest answer?
Bless: We won't know unless we ask patients. We also have to pass on other questions if we forego the patient's perspective, for example how effects and side effects should be offset. Or how to include the disease-related quality of life - which itself already includes dimensions that describe morbidity - in the balancing without redundancies. But the procedure provides for an aggregated overall measure, even for a subpopulation. Therefore, such a balancing decision must always be made, but at the moment it is exercised in a paternalistic manner.

Incidentally, decisions are also made paternalistically about what is relevant to the patient and what is not.
Bless: Exactly. This applies in particular to the so-called surrogates such as the time of progression-free survival in cancer or the durable virological response (SVR) in viral diseases.

IQWiG repeatedly explains that this has not been methodologically validated and is therefore not relevant to patients.
Bless: This approach, which is reduced to methodology, bothers me. On paper, the patient is placed at the center of the procedure, but it is third parties who decide what is important to the patient and what is not, and how he or she has to decide and what and how to evaluate it.

Your opinion on AHP and CA?
Bless: I think that analyzes that can be carried out with these instruments can also be used in the early benefit assessment and should be used now in order to be able to make exactly those balancing decisions much more validly. Because so far, as Mr. Müller explained, there is no recognizable methodology for offsetting, which, on top of that and in contrast to the otherwise very transparent behavior of the G-BA, is carried out in an absolutely non-transparent manner.

Would the solutions that IQWiG has brought into play be a viable approach?
Bless: The IQWiG at least has a calculation algorithm for the benefits, but here too the balancing remains a hidden balancing decision. But the G-BA did not approach this calculation algorithm anyway, but instead opened up the leeway that the wording of the legal regulation allows. I wouldn't describe the terms as semi-quantitative, they are semantic tautologies. What exactly does it mean if “minor” is defined as “moderate but not insignificant”, what use is it if “considerable” is explained as “significant” or “significant” as “major improvement”? This inaccuracy actually gives any leeway whether an effect measure is moderate, small, marginal, or whatever. There is quite a bit of scope there, but that doesn't have to be a bad thing per se.

Why?
Bless: Because this allows the G-BA to take into account other aspects than a pure calculation algorithm can achieve. But where balancing is done in the name of the patient, I would like the patient to be included and their voice heard by questioning them - and the methodology is sophisticated enough for that.

Müller: The legislature has provided for very intensive patient participation in the G-BA. There are both topic-related patient representatives and permanent ones, all of whom work very intensively in the AMNOG process. But I have to warn and slow down with regard to the models discussed here: The legislature demands an evidence-based decision in the AMNOG. He refers to scientific evidence and not to patient preference. What they do is partly pure opinion research. And therefore something completely different than evidence-based medicine with effects that can be seen from valid studies.

This is called a conflict.
Müller: Exactly. The G-BA must make an evidence-based decision in which patient preference can play a role, but not the decisive one. Precisely because it is more of an opinion and not a scientific result. As a doctor, I am skeptical as to whether majority decisions can actually be made here. I agree with you that patient preference can be used to better capture certain effects, such as spasticity in MS therapy, which can be very restrictive for the patient in everyday life. If, with the help of such analyses, such effects can be recorded reproducibly and therefore more validly, this is also relevant for the Federal Committee. Such findings cannot, to use the word, be dismissed paternalistically. But once again: the evaluation and the formation of opinions in themselves do not meet the tough requirements of evidence-based medicine. The G-BA must follow a clear line.

Then the much-vaunted “learning system” AMNOG must learn to deal appropriately with such economic models, which already come with a high level of scientific professionalism. If I were to confront the professors commissioned by IQWiG with pilot studies on AHP and CA with the term “opinion research” that you, Mr. Müller, use, they would jump.
Müller: If validated procedures such as conjoint analyses, discrete choice experiments or the Analytic Hierarchic Process can provide information that corresponds to that of evidence-based medicine, they should be used. In the cases described in which positive and negative effects contrast, the G-BA carries out a balancing based on value estimates. This can also lead to majority decisions, even if these have so far been extremely rare - the vast majority of decisions have so far been made unanimously and with the consent of the patient representatives. This decision then applies to the respective patient group and is the basis for price negotiations. However, the individual doctor and his patient are not bound by this, but must make their own decision.

Bless: The central methodological challenge is always the selection of a representative survey group. Depending on the task, these can be patients affected by a specific illness, or a community of insured people, because these are the ones who have to bear the costs. The questions must also be selected in such a way that no overweighting of individual aspects creates a bias because they are redundant and are practically included several times. This requires a fine feeling and a lot of experience. The methodology itself is not the problem, because it is now very sophisticated because it has long been used beyond the healthcare system to ultimately determine people's willingness to pay for individual attributes of products.

Typically with cars.
Bless: Or for eco-labels on the washing machine, there are even discrete choice experiments for that. But that means nothing other than that these instruments have been validated for a long time and are now the gold standard. Just not in the health sector.

But how do you ensure that a so-called theoretical opinion is not asked for?
Bless: When an illness occurs, those affected are initially selected because only they are able to make a value decision. This is much more difficult for a healthy person who has to do this in the abstract. The questions are also validated internally and careful attention is paid to ensuring that there are no redundancies so that the weighting is correct afterwards. There are also preliminary pilot stages in which small panels are surveyed to see whether the selected questions are suitable for the method used in order to obtain usable statements. Only then is the study carried out and the results statistically evaluated. This is designed in such a way that the individual attributes are given an evaluation and so the overall measure of various aspects can also be evaluated.

Müller: The challenge for the G-BA is to take not only the patient perspective but also the insured person perspective. We experience this conflict in almost every drug evaluation, in that the topic-related patient representatives look at certain effects from their own perspective and clearly express this opinion - also in a very understandable way - for example on mild hyperglycemia in diabetes. Then it is actually not an easy task to find an overall perspective in a body based on plurality like the G-BA. In this respect, the criteria specified by the legislator in the Drug Benefits Ordinance are very helpful. The division into mild, moderate and severe side effects may be a bit rough, but after a differentiated look at the first 80 AMNOG results you can clearly see that there are big differences between the products. And they are also shown.

They said there is no single true cost-effectiveness assessment, and certainly no overall benefit assessment across different drug therapies. How are you ever going to get there?
Müller: That's a good question. My opinion is that we still need a comprehensive discussion on this.

We just want to encourage them.
Müller: This is also necessary if we consider the current development in drug prices, where some orphans have already reached annual therapy cost “sound limits” of up to one million euros. I am convinced that we also need a methodical approach in Germany in the medium term in order to achieve a fair balance between costs and benefits and prices. In my opinion, however, more economists - and not just health economists - and their tools should be involved.

Why not just health economists?
Müller: Because the question of a fair price is not just about QALYs and efficiency limits, but also about economic questions. Discussions will also need to be held here in the future to create a balance between price and expenses for studies, but also incentives should be given for research and development. And at the same time, too much money should not be withdrawn from the statutory health insurance system so that it does not become unstable.

Wouldn’t the QALY concept be a tried and tested approach?
Müller: I think neither the QALY concept nor the efficiency frontier is sufficient. Both methods lack an overall economic perspective, and both methods do not relieve the legislator of the value decision on the question of what a fair and acceptable price is. To do this, the legislature would first have to take the next, much more advanced steps, but this is not foreseeable. At the moment I can't even see that we're getting into the cost-benefit assessment. After all, we just had the news that the 50th reimbursement amount negotiation was successfully completed. In this respect, the legislature's approach is currently quite successful.

Bless: I would also predict that the legislature will continue to wait and see what the self-government does with the current framework. That has always been the case, and there is currently no urgent reason to abandon the principle installed with AMNOG. I would also like to question whether you really need an overall measure of benefit across indications. From a purely benefit perspective, I don't have to compare the healing of a middle ear infection with the quality of life of oncological patients. Because this type of information is meaningless from the perspective of utility, but not from that of pricing. I would also like to question whether we should long for the QALY concept, because we quickly reach the question of what a year of life should cost.

This is a question that will certainly be very difficult to address in Germany anyway.
Bless: But if so, it should be conducted in society and certainly not by HTA authorities. We can already see in the method paper of the IQWiG cost-benefit assessment that there is a small, remarkable chapter called “Reasonableness of the assumption of costs”. This is still very reserved, but ultimately the ultimate consequence is addressed every time a price is calculated: What if the calculation result means that the GKV cannot cover the expenses? Once this final consequence occurs, we will enter a new era of allocation decisions. Then an ethical value debate begins, which I would not like to see located at IQWiG or the G-BA, but would always anchor it in society. Therefore, the negotiated solution pursued so far is probably not the worst, because the G-BA presents the evidence in relation to the benefits. Of course, the negotiating partners can take all sorts of other aspects into account, be it the supply offer, the supply requirement and other effects.

But there are also completely other aspects that determine the price and that influence costs.
Bless: Secure. When setting prices, the number of patients served is important first and foremost. I assume that the development costs of a drug will be relatively comparable, regardless of whether it is used to treat 100 patients or 100,000 patients.

But what is the consequence?
Bless: In a small group of patients, the new drug has high incremental costs and therefore poor cost-effectiveness. In the logic of QALY, this can lead to people with rare diseases no longer being cared for because pharmaceutical companies cannot price in the costs that they actually need. NICE has recognized this and is therefore abandoning the cross-indication assessment for Orphans.

Müller: I see this problem of QALY and the development in the UK in the same way.

Bless: As far as costs are concerned, we have currently looked at the pure cost bearer perspective and only the so-called direct (medical) costs.

But what about the social costs?
Bless: There are indications - I'll take rheumatism as an example - in which a large part of the financial burden of the disease falls on society or other payers. This is due to incapacity to work, retirement and care. It is a very difficult situation when statutory health insurance is supposed to set a fair price, although it actually does not benefit from the financial benefits because the savings are located somewhere else entirely. But that is exactly what would be provided for in the cost-benefit assessment, that the perspective of indirect costs can be included at the request of the G-BA.

Has this been done before?
Bless: No.

Müller: Not yet. Ultimately, this is how the legislature designed it, but the vast majority of procedures end with price negotiations. The stage we are talking about here would come after the price negotiation. But neither an entrepreneur nor the GKV-SV has applied for this so far.

What is the level of satisfaction with QALY in England?
Müller: There are paradoxical effects. For example, new products in oncology often enable a few months more life expectancy, sometimes associated with a poorer quality of life. This then leads to the products in the QALY concept being priced so high that the financial limit set in England is exceeded. This in turn led politicians in England to set up a supplementary “oncology fund”. In addition, an “end-of-life regulation” was implemented some time ago, which extends the QALY limit in the oncology area. But in general it must be clearly stated that the QALY concept means that some new medicines are not available to the population. In Germany there are no restrictions at all in oncology. We do not remove any products from the market for cost reasons, whereas this is certainly the case in the countries that use QALY.

But that doesn't seem to trigger much social resistance.
Müller: I can't say for sure, but I assume that there are discussions about this in England and that the new government has set up an oncology fund for precisely this reason. There has just been a publication showing that the average median survival times for cancer patients in England are worse than those in Europe. I wouldn't want to claim that this is related to the QALY, but it is clear that in oncology the QALY concept leads to restrictions, while in England, interestingly enough, in other areas such as antibiotic administration or hepatitis therapy high costs are accepted when it comes to many years of life gained with a high quality of life. At least the NICE reviews are quite positive in these areas.

Bless: We have just carried out an internal comparative analysis in oncology at the IGES Institute. We compared what constitutes the G-BA and what constitutes the NICE ruling practice. I found this in-depth look important because some people always long for NICE. Overall, it can be seen that the benefit assessment of NICE is less rigid than that of IQWiG. The issue of surrogates is also dealt with in a more relaxed manner. That means: In general, benefits are recognized more quickly in England. But then comes the hard threshold of economic consideration, through which some products do not reach the supply. This affects a number of products that the G-BA has recognized as having considerable additional benefit. On balance, we have seen that the German principle ensures that oncological products actually reach the patient much more than is the case in England.

What about social acceptance?
Bless: I assume that a lack of acceptance has led NICE to violate the QALY principle through the oncology fund or in the area of orphan drugs. The QALY principle actually knows no exceptions, as it aims to be valid across all indications.

Still, the QALY system rewards innovation up to a certain level, like 70,000 or 100,000 pounds or dollars or whatever. We only ever discuss a discount.
Müller: Let's take an appendectomy: A ten-year-old patient comes to a good surgeon early and lives to be 80 years old. In the QALY system, 70 years of life gained would then be multiplied by an assumed value of £40,000, which the surgeon would then be entitled to. The example shows well that the QALY methodology can lead to misleading results when setting prices. Medicines for hepatitis therapy are rated so positively by NICE because patients are cured.

You said cured?
Müller: I know that people can have different opinions on this. If the virus is no longer detectable, liver cirrhosis does not develop and a liver transplant is not required, one can speak of healing. However, the sustainability of the effect still remains to be seen. If you extrapolate this to the years of good life gained, the QALY concept rewards this innovation. This is different in oncology, where a patient often only has a very short lifespan and a poor quality of life. Here the calculation looks completely different: around 1 month times 40,000 pounds times 0.001. If I now put this price in relation to the price of the drug, then these products perform rather poorly in the QALY system. That's why you can't say across the board that innovations perform better or worse in one system or the other. The comparison with Germany is also difficult because there is still no social discussion or even a consensus about where a possible border would be.

Mr. Bleß, in your article “What has been learned and what is still to be learned” in “Market Access & Health Policy” (issue 05/14) you wrote that the G-BA and the IQWiG are unfamiliar with a balancing of interests. The AMNOG process is – as it says in this article – “primarily aimed at avoiding type I errors – i.e. false positive evaluations. However, errors of the second type - i.e. additional benefits that are wrongly denied - are accepted here." This particularly affects manufacturers who are faced with the situation that the studies that were agreed with the approval authority are not available before the IQWiG and the G‑BA take into account. An additional benefit that is wrongly denied can certainly have negative consequences for patients and should therefore also be urgently avoided.
Bless: By applying the rules of evidence-based medicine, care is taken to ensure that no false positive statements are made.

In other words, attributing a benefit to a product that it then doesn't actually have.
Bless: Exactly. But when I look at the decision-making criteria of approval authorities, there is even more on the table. On the one hand, the approval authority also wants evidence, including patient-relevant endpoints, but it always also considers the existing medical need - i.e. the need for an innovation in a specific indication. The regulatory authorities are thus asking themselves how it should be assessed that a new drug is made available to the population. This culminates in the conditional approval, which officially states that the lack of evidence is to be considered lower than the population's need for this new drug.

Here, too, there is a balancing of interests between two values.
Bless: Absolutely. But what I'm getting at is that, in contrast to G-BA and IQWiG, the approval process recognizes this balancing of interests. In the area of conditional approval, the G-BA has also used its leeway to avoid contradicting other authorities, but this is alien to the procedure itself. Likewise, patient-relevant harm could be caused if the G-BA denies an additional benefit and a drug is then taken off the market, for example. Sometimes I would like the process to admit that it can be wrong. There is no other way to do this, which is due to the low certainty of results. If I have an indication of a benefit, then the possibility of error is inherent. And you always have to weigh up what damage I can do if I don't provide this additional benefit. But I miss this balancing decision in this entire process.

Müller: This results from the different tasks. Approval is primarily about product safety and consumer protection. It is important to avoid products coming onto the market that are not safe and of poor quality. When it comes to effectiveness endpoints, i.e. the positive effects on the therapy, the approval authority sometimes sets lower hurdles than the Federal Committee in its benefit assessment. The G-BA checks whether new products are better than the standard. According to the legislature, the G-BA must make evidence-based decisions and, with regard to the solidarity-funded statutory health insurance system, ensure that what is necessary is not exceeded. The licensing authority cannot fulfill this task with its completely different legal framework.

Mr Bleß and Müller, thank you very much for the interview.
The interview was conducted by MVF editor Prof. Dr. Reinhold Roski and MVF editor-in-chief Peter Stegmaier.

Citation information: doi: 10.24945/MVF.05.14.1866-0533.1958