Early Development and The Dose Selection in the Immune-oncology

Early Development and The Dose Selection in the Immune-oncology

Alexander: [00:00:00] Welcome to another episode of The Effective Statistician. Today I’m super happy to have Teppo with me. Welcome to the show. Maybe you can introduce yourself shortly to the audience.

Teppo: Thank you, Alexander. Hello. Hello, everyone. So I’m Teppo Hutturen. I’m a statistician and I come from Finland in Turku so southwest Finland or 50 kilometers from Helsinki.

I did my statistics studies in the Turku University graduated 2004. And from then onwards, I’ve been working with with the truck development and a long time in a Finnish CRO. And then three and a half years ago me and a couple other colleagues we started our.

On company estimates, which we have been now running and we have been now developing the company and we have a group of 10 experts there. We also have data managers. We [00:01:00] have programmers. We have also a small data science unit. But But me and then many people, we have a background in, in statistics and in, in the biostatistics.

So that was where my career started when we had a, in Turku university, there was a master’s program focusing on the drug development statistics.

Alexander: Yeah. So you’re basically a boutique CRO. What are your typical clients that you look for?

Teppo: We are Collaborating a lot with the biotechs the biotechs don’t usually have their own statisticians, then we can operate as their internal statisticians and be the statistical experts there.

Then we can conduct like also the operative part, but then many occasions we are doing there might be a CRO involved and we are doing like the statistical oversight. That is typical. We also serve some clients in [00:02:00] the functional food. So there we are more doing the operative role.

Why? In, in the food studies, but then also with different kinds of pharma companies, there might be early phase studies like DK studies and these kind of studies from different sectors.

Alexander: Okay, cool. Very good. So today we want to talk specifically about oncology and especially immune oncology.

And one of the first challenges that you We’ll get into when you develop these tracks is of course, that those selection and I’m not an oncology expert, but one of the things that I know about is that lots of these oncology early phase studies where it’s about those selections have says three plus three design.

You basically start with three patients. You look Jose turn out to be, and then [00:03:00] you escalate further until you reach some kind of maximum tolerable dose. And that’s basically your dose finding. Now, it is known that this design has many different disadvantages. What are a couple of the disadvantages of these studies?

Teppo: Yeah like you mentioned it’s really commonly and historically used methods and basically it is a very effective method when there is a cytotoxic agent. In question. Yeah. And then it’s purely based on the hypothesis that what is the most tolerated amount of the truck is also the most effective.

And then Only with three patients or then if it’s the three plus three, six patients you can get this information with quite high probability that, okay, you [00:04:00] will eventually run into this maximum. tolerated dose. And then you have the dose for your phase two study. But then, now quite a long time already, we are being, we have been working with the immuno oncology molecules.

And there it is not necessarily that, that the. The maximum tolerated dose is the most effective dose. Okay. And I think this basic question is like what the companies and the people working with the immuno oncology and in the oncology drug development should understand.

And then when you understand this, like a paradigm or this issue so And then you can start to think that, okay, then how do we set the question and questions that we would get the better better estimation for what could be the the optimal biologic [00:05:00] dose.

Alexander: Yeah. So that’s really a paradigm shift from looking into just the maximum tolerable dose to the most appropriate dose in independent or not independent, of course, as a kind of a upper limit, but it doesn’t need to be this kind of upper limit.

Teppo: You need to, basically, I think it’s good if you can find it, the limit, because that answers the question then that, okay, what? are the safe doses. Yeah. But then the next question should be that, okay, what are the most probably efficacious doses? Yeah. Yeah. And then you should widen the thinking and yeah, and I think there then what is, and I think that, Authorities are, of course, aware of this and they have been FDA has raised different guidances and there is [00:06:00] nowadays also a kind of fit for purpose stamp on this BOIN method.

Alexander: What does BOIN stand for?

Teppo: That’s the Bayesian Optimal Interval Design.

Alexander: Okay.

Teppo: Yeah but basically Boeing is it’s like a more general it has the same kind of utilities that the three plus three, basically three plus three is a special case of included in the Boeing.

Alexander: Okay.

Teppo: So actually the bone is also answering to the same questions that, okay, what is the maximum tolerated dose?

So we are still in, in the same framework but there is a lot more flexibility with the bone.

Alexander: How do you get, how do we get to this more flexibility? What are the other parameters in the introduce?

Teppo: In the boy, and there, there is a lot of modeling behind, but basically there is you set [00:07:00] up a target toxicity level.

It’s also in the three plus three, you are having the target toxicity level basically of 30%. In the Boin, you can set up the target toxicity level also to 25 or 20. In case you want to like not allow more toxic Agents and then you have a decision rules for the dose escalation and dose de escalation, which are not tied to the number of patients.

So you can have three patients as a cohort size and then your decision rules are, patients. Pretty equal to the three plus three or in some cases, even equal, but you can also have to call her choice of four or five. And then it allows that if there are cases where the. Patients discontinue early and they are not [00:08:00] like a DLD, the dose limiting toxicity, evaluable.

Then you can still make decisions about whether you need to stay on the same dose level or you need to de escalate or can you escalate to the next level.

Alexander: So basically you can have more flexibility first to understand what is your maximum tolerated or your maximum toxicity set that you want to get below.

Teppo: You are more flexibly moving. We along with the doses . And then the operational characteristics are built so that you have more patients allocated to the to the doses around. Or below the maximum tolerate those because the three plus three is basically running that you run to the highest level and then you stop.

But with the Boeing, you can based on the observations, you are moving more, more around the the MDD and you get more [00:09:00] information then also about other parameters than only the tolerability.

Alexander: What are these other parameters that you can measure?

Teppo: It’s the responses. It can be also the like PK and PD parameters.

Okay. And also the utter tolerability. Because these those escalation methods are based on the dichotomic that if there is a DLT or not, but then of course you can have like more, more milder adverse events.

Alexander: Yeah. Yeah. And so you can also take an, and here’s the responses.

So that’s basically the efficacy also comes into play. And you can basically also model that, so it takes that into account. So that you also get information about the efficacy, not just the tolerability.

Teppo: Yeah. There are also these kinds of models that you You can use the same kind [00:10:00] of framework and accounting for the efficacy responses.

 And these are called, these call so called toxicity and efficacy models. But there I think there I haven’t seen those that much in the practice because already running the model Using the DLT events, it’s, it starts to be like the prac somewhat of a practical challenge.

And then if you need to follow up also the efficacy responses and include those in the decisions it starts to get complicated.