Explore the Scientific R&D Platform

Free Trial

Impact of COVID-19 in clinical trials [webinar]

About the webinar

In this webinar, guest presenter Dr. Luis Rojas, former Senior Principal Statistician from Parexel, now Executive Director Head of Biostatistics at Target Health examined the impact of COVID-19 in clinical trials. You can the watch the recording of the webinar below.

The Impact of COVID-19 in Clinical Trials

Guest Presenter:
Dr. Luis Rojas
Former Senior Principal Statistician from Parexel
Now Executive Director Head of Biostatistics at Target Health


In this free webinar you will learn about:

  • Understanding the impact of Covid-19 in the areas of:
    • Clinical Operations
    • Data
    • Statistical Perspective

  • Composite intercurrent events vs ICE categories

  • Statistical Covid-19 strategies including:
    • Study Design
    • Reporting
    • Measuring the effect

About the speaker
Dr. Luis Rojas
Senior Principal Statistician from ParexelLuis-Rojas-nQuery-Impact-of-COVID-19-In-Clinical-Trials-By-Luis-Rojas-slideshare-webinar-slides

Play the video below to watch
the complete recording of this webinar

Duration: 60 minutes 
Nothing showing? Click here to accept marketing cookies


Looking for more resources?

Transcript of webinar
*Please note, this is auto generated, some spelling and grammatical differences may occur* 


Today's webinar is The Impact of COVID 19 in Clinical Trials. My name is Ronan Fitzpatrick, the Head of Statistics here at Statsols. But today we have a very special guest who will be presenting today's webinar, Luis Rojas from Parexel. It's a great pleasure to have you here today. Luis, very much looking forward to today's webinar.

Thank you Ronan. So, hi everybody. As said before, my name is Luis Rojas I am a biostatistician at Parexel. As we get started let me read the following disclaimer.
The views and opinions expressed are those of the presenter and should not be understood or quoted as being made on behalf or reflecting the position of the international conference on harmonization, any regulatory agency or Parexel.

Today, we're going to be talking about and provide some evidence to help you understand the impact of COVID-19 in clinical trials.

And then we're going to navigate to the differences between encompasses intercurrent events and different intercurrent events. In this particular case, we are going to be doing an evaluation when we consider a co-ordinated team as a composite endpoint.

And when we break it apart in different intercurrent events, we'll talk about covid- 19.

And when you consider each possible category associated with copying a team that's at these phone, into current events. We're gonna be talking about a statistical considerations for studies, and perspective, From a reporting perspective, And actually, at the very end, we're going to be talking about how to measure the effect.

Opening remark in this era, when, in the last few years, probably since 2007, being more heavily last year, with their, within creation of the final guidance for the esteem and IC, HP 9 R 1. You may be familiar with it for other ones.

But, back in the day, we used to refer one of the key task during protocol development or research in general, as the operationalization of the research question. Why we're doing the research on what is the justification of doing what we're doing, we're about to do, and then what needs to be measure.

All the dependent variables get into the independent variables and the dependent, again, all the factors should be considered in the evaluation of their primary secondary endpoints, and then intercurrent events, any issue that may impact our evaluation.

In the past, we didn't use the term intercurrent events. And in the limitation in this study, we will define under which conditions on how each condition that may interfere with the evolution of the primary endpoint will be evaluated.

Also, a population level summary. Basically how the data will be analyzed, and then all the treatment intervention is the recent treatment information about yourself. 

That's something that only happen in intervention studies, but it could happen and it should be happening. Also, in observational studies as well.
Very well defined and point very well defined injective. And then the when the research will be done in the tie lines, When the interim analysis are going to be executed for simple when DMC's are going to be executed as of ours on.
And then, where did your region, what is the location of the body, for example in where the measurements are going to take place? We're talking about the estimate. Think about each region, this is impacted differently by COVID 19.
When we are trying to measure, our endpoint that we will have a different impact in every region in relationship to another one. And you will see later how we're gonna link all these concepts, will be our population of interest. And, again, we'll see the link between the Population of Interest and Covid-19. Then how we handle the discrepancies in our original patient population.
Many of us are already trying to work with the data was collected. Is there any change in how the data was collected in? The original plan was about to be collected injury and a plan, and how the data was collected, after a common theme, is the interaction between the physician and the patient the same. Or are we use a virtual visit, remote, visit, or whatsoever, so far song, And then the analytical method. Again, associated with a population level summary, and all the changes, the validation process for not only the analytical method, but also for the instrument.
Then, finally, we control, for my ability to manage the placebo effect. How are we going to handle missing data, which, again, is highly associated with the formulation.
And then how are we going to handle dropout cross treatment issues, rescue, medication, discontinuation, which at the end?
They need of operationalization of the research question which will begin to do cortical development.
As you know, an estimation before the estimate, we have a, our current definition, which we do respect. I have my personal opinion. I don't think that is a precise description of the treatment effect.
However, it gives you some evidence of possible understanding of the treatment effect.
I do, I prefer to use evidence of treatment effect instead of precise description, because we're taking the estimated from the sample population.
And then those by the clinical question, by the trial objective, summarize at a population level this level, we're going to be in the same patient in the same, based on their treatment condition being compared. And I said the same patient, because ... basically is just in the formulation of interest.
We have patient impacted in different ways and handled in different ways during the life cycle of the trial.
I want to talk about the estimate of the estimator.
So you can compare the current definition of the estimate with that the way we should do it in the past, and they still done in many areas in clinical research, which is called the operationalization of the research question very important then maybe new persona for you. I call it the CWA effect.
I saw the terminology, was explicitly mine, but I think that is, we use in the bus very, especially in psychology research.
But, I will be referring to the Civil website, to the combination of the placebo effect that is a passage at the station coming in from the subject or the participant in the clinical trials. But, also, then, see, what effect that is the negative expectations about the effects that the treatment effect. So, I like using the term Siebel Effect because not everything is glass Siebel. There are based on or subject who may have a negative expectations.
And they will diminish the interpretation or even the will diminish, the treatment effect because of interests.
In fact, again, that's where the copywriting had where we see the receive website, if you use them by intrinsic factor, streets or factor, I would like to go a little bit about all of these items, neurobiological, psychological, and cognitive factors. Lobby and conditioning the hub horn effect the experimental subordination, the natural disease progression, probation to demean the disease status of baseline. So what do you think?
Oh, well, all the, all, the possible issues associated with the person, or the patient itself, The relationship of the patient, with this site, with their clinicians, the characteristic of the evolution of the, the, the particular disease under investigation, and how the relationship and heart disease, actually changes over time.
And so when you add the co relating issue, then you need to take them. So we need to take into account very specially, that, in many cases, in many clinical trials.
This setting in which the original study was plan, is changing.
two, in some cases, virtual relationship between the patient, the primary investigator, or the investigational team. Sometimes you have been completed, VC that I've done over the phone.
You may have be a conference where the clinician sometimes there are restrictions to go to the site and then suddenly testings are done on that, what we call now, virtual clinical trial.
So, the expectation of the effect size cannot be the same and will never be the same.
When you have innovation actually move into a site, preparing himself or herself psychologically to attend a visit.
Doesn't know when the patient has been in home at home for a very long period, as many of us, And then having to deal with this with the BI, or the CRA over the phone or via the internet.
So, instead of cost, or I refer to those associated with environmental factor that metal the methodologies for quantifying the treatment effect, background medication on social interaction, and here, social interaction are key.
Because, we actually know, the reset proportion of the treatment effect induce, by the fact that you are dealing with, with people, who may or may not have the same disease. But, when the social interaction is limited to just, you, are confined to your home, because of the 19, then, both interests and electricity factor by going to be slightly different.
OK, now, how can we, in 18 85 clinical trials, I seen amazing papers published my colleagues around the industry, but I would like to give a slightly different classification that covenant chain can impact clinical operation The data yourself and then for an a statistical perspective, they are going to be also an impact lecture for subclinical operation, site selection, cyber retention recruitment rate.
Are slowing down in many places, because of going I T, So that is almost impossible to retain a site, because the whole site, or the whole region is being shut down. And then they are restriction for mobilization of patients. Are participants within that particular region. And then, the way this site is operating is completely different nowadays, and especially with implementation of something that we call virtual monitoring, or remote monitoring. And then dropout rate is also increasing, in some cases, not always.
But in some cases, increasing buddy's also depend on the nature of the clinical trial.
That's going to be sharing this information is something that I would like to order is this is not a solution for all of the clinical trial. You just, in general guidance that I'm trying to share because it changed from study to study and oncology trial executed at a clinic will have different empires. That a trial for a suitable for musical are nice trophy in the pediatric population in where you do have to take to do a site to do a physical evaluation. So these two cities are completely different.
For a sample of some clinical trial executed for dementia, or any psychiatric disorder, or any CNS disorder, will have a different connotation. Because then you need to add to the current indices. You need to add the fact that the patient is being confined at home, or may or may not have extreme limitation to socialize with other people.
Bernadette perspective. Data collection is getting more difficult, right? We are, in, like I said, we are adding or removing site, and then, or we are adding or removing. Third bodies vendors who are processing the Data Flow symbol for ECG for labs, Beta some even to places where they run the labs. They may be also impacted by co-ordinating, and then you need to add in additional resource to process the same information. And that makes things more complicated and collecting that data, and integrating that data becomes more of a challenge. That issue, as far. As I mentioned, I am a Biostatistician by excels Plus, or make sure before I leave the situation in a daily basis.
Easier graphs are getting updated, because we're trying to accommodate what is happening in reality, due to the needs of the clinical trials and dose modification to measure the impact of going ID and, or even to measure the seas or measure measure. Their procedures planet in the clinical trial are changing, because we need to adapt to the changes seen that in the environment, make.
Clinical data and the processing of clinical data more difficult.
Data monitoring and data replication. Again, adding or sample virtual monitoring and verification processes is different.
Like I said before, the integration of the data is more challenging.
And, of course, we need to keep our eyes on all our effort to ensure that the data, the quality of the data, is optimal to ensure consistency within the trial.
Finally, compliance rate compliance with the study procedures, Compliance with the treatment administration and it will become more difficult.
Why do you have patients that can now have access to the site or this site has changed the procedures for administration of the therapies or for the measurement of other endpoints and then the compliance is compromise granted physical perspective.
Of course, we have patients being a drop in front of a trial or patient impacted by Coble ID, and then we either lose power or if we have an interim analysis where we are trying to compete, For example, condition of power.
Get also compromised where we're trying to predict the probability for success on any given time. BJ, compromised, multiple, multiple example.
So if you're planning a DMC or you're planning an interim analysis, for example, 30% of the data on the time he's running, Recruitment is really slow, and then it is a time to run the instrumentalist. Everybody, instead of 30%, you may have 50% of the data. And then how to handle the situation. You're gonna be performing the DMC other particular particular event. Time points because it's time to shake the safety profile of those patients.
Who are more in, the, few patients are moving towards the experimental trial Or should we wait for additional patient to complete the sample size Required For the, for the DMC? So it is a challenge. Most of the time we're going to have to, Run DMC will lower sample size because we need to ensure the safety of the patient is monitor accordingly. So, that, change the rules and how are we going to compute, for example, condition of power.
So then you do good how, the evaluation of safety. And then you're going to have to wait until the recruitment increase. And you get more some more based on randomized to perform the evaluation associated with condition of power, if you're doing a sample survey estimation, for example. But it is not recommended to reduce the sample site is basically, or conditional power.
If you're making that prediction, that particular given point because of the variability in the respondents, most likely you're going to increase the effect size on the CEBO effect.
Like I said before, and now, you can see, the relationship between my previous slide, and this one, the settings are changing.
The second set of changes and base shark on flying in their homes, like many of us, are like I said before, and then, this, might need to, of a placebo response.
Do they know CBRE responds in general is totally different than the assumption of May?
Before the trial started where we were designing the clinical trial?
me as a, as an expert in trial design?
I find it very difficult to assume, or two, to rely on data about the effect size that West competed before the Company 19, because ... compete in many, many clinical trial work period.
We said, in a normal condition, the West competed with patients who are not facing Kobi 19 patients who were not diagnosed with golf club in 18.
Then the effect size is going to be totally different in this circumstance, especially in immunotherapies, where you have many of the cases you may have a delay effect, or you may have a diminishing effect, or you may have a crossover issue, for the hazard ratio.
Picture originated and increased. So this specter variability, I think, that I mentioned before, and is highly related with the sample size reminder, when the sample size is smaller than variability, may increase.
And when we change site A site are dropping or the number of subjects per site, changing from the original plan.
And then, the variability will also increase, as we all know, Because some of these responses will or could be associated with, geographically regions, with some details associated with the patient characteristics, or the treatment of ministries tuning in certain regions. Region is one of the risk factors associated with the respond positively.
So, that needs to be taken into account sample's second replacement.
Uh, I'm not 100% sure if my peers are all also thinking about the same. Pretty sure. Many of us are in the same page with a sample site replacement, especially in biosimilars, and by equivalence trial where we have to use the protocols that we used to. We need to use Bayesian completers to ensure that we can capture the became profile in an optimal way.
So in many cases, we are forced to assemble a second replacement and ensure that we are to end the demolition of the primary endpoint use in completers instead of patient randomized which she said, a major deviation from the intent to treat the principles. But I think that would be, justice, a case, by case basis. You need to be discussed with the regulators.
Have an agreement, what can I do if my trial cannot use the intent to treat? Principle, I may end up using the modify. Full analysis said, because they call the ...
site, being close, patient drop, or like, the compliance rate is much smaller than expected than my personal opinion. Will the boat and I will reach that assemble site replacement needs to take place? Planted the fine estimates.
If you know the estimate pretty durian culminated in a completely different somebody yesterday was asking me, what about this? truly, try death, Sidney Dekker an event with a different connotation now that we are in the repressor of Cabinet deem. My answer was very complex in nature to certain extent.
I was suggesting, Yes. Absolutely. Yes. Because he's not saying the cost of that, the cost of it is to combine a team and it is a confirmed case recombinant team.
Then, you shouldn't be associated in the response to co relating to the treatment effect, but, to cover anything that could be done by an in app, the natural history of the uses, that could be caused by multiple research, and each resource for bed treatment. Emergent adverse event that we need to be evaluated using a different inter cooling strategy, and we'll talk about that one layer, statistical method. We'll have a slide for this one sensitivity, seulement and analysis, and then the interpretation of the results will also be document in a little bit.
The impact of clinical data in the Cabinet deem as a composite into a current event versus considering COVID 19 separate categories, we super subcategory, and then treating each category its own intercurrent events.
I said a big category as a composite intercurrent events.
Do they find it couldn't have any ties, because they will be multiple implication and especially if you have different at three minutes depending on the trial Was two years statistician's and the medical team sit together and define what are the implication of calling a team in in at least element that. I have a slight compliance.
Treatment interruption, delay pretty much for treatment discontinuation.
These are prohibited medication or rescue medications because of copying a team that is different, when you use a therapy for copying a team.
But that therapy and be prohibitive medication for your particular trial is different than when baby should use a prohibited medication or a rescue medication because other recent ..., so rescan therapies could have a completely different meaning in in in different end points.
So a rescue therapy could be prohibit the medication in the region of Atlantis, buddy, finisher, rescue therapy, induce by an issuer costs by copying a dean. Then, the inter cooling strategies need to be accommodated to ensure that is handle accordingly.
See, there's hospitalizations use of respirators, whatsoever, like I said that due to cover 19 hospitalization due to COVID-19.
So, you can see when you add a lot of complexity in our clinical trials it is strongly recommended that we treat each one of the into current events.
Impacted or induced by Coburn a team as its own entity.
That's one of my recommendation, then how we handle it. Remember, this is not a saloon.
Not each integrated strategy will be applicable for multiple studies are different indication. At the same time.
You will see that I have some in bold, and they are highlighted in this way right there with the line at the bottom. That mean that these are the one who are not highly recommended. And the one that I recommend that are the one without the actual line treatment policy strategy, where we consider that a corridor in your relevant in defining the treatment effect.
Dad, through that integral around strategy, most likely would not be, of, scientific interests, for most covenant chain related intercurrent events because, of their conclusion will not be will not be able to generalize their conclusion in the absence of the pandemic.
They hypothetical strategy in which the interest is in the treatment effect intercooler ND not occur.
Uh, I recommend personality to do a mixture of cases who adhere to the treatment doe, who do not here, to the treatment, because of their ....
Some cases, you can go across symbol, assuming that all the patients, you know, all the ambition impacted by ..., could be treated.
and even if the data is collected, I see in that approach something into trial where all bases or all data from business impact by copying a team are treated as missing, and then we do a multiple imputation assuming at random.
They are some of my peers in the industry.
There are also suggestion and indicate in that, if you're planning to use myths missing completely at random, be careful, second assumption.
Before proceeding, in some cases, that may be the case, but not for all intercurrent events. You can justify that is completed at random, so an offer of regional for this continuation. So, we have to be a one-on-one evaluation, and check your particular case. But technically, astronomy, in general, is this is a valid approach, can provide some different rationale of what happened, for those cases, where the data was collected, or when the data was not collected, or when the data was collected. And the team decide, no less, assumed that all these data was missing at random.
Principal stratification justifications strategy.
Not recommended.
I don't recommend using education buy any fact, or any intercurrent events associated with probability.
I remember at the beginning of a pandemic we're seeing to, well, what, if we had 25 do?
We add Gobi 19 I circled areas, and it was too early in the game to understand the actual impact community. Now that we have a better understanding, the recommendation is do not try to stratify or do ... strategy, because the interpretation that the result is going to be completely different.
And the S too many cells that you're moving towards, that you are trying to compute right, to the estimation procedure that you have in place, will not be accurate and will not be associated with your primary endpoint. So careful consideration when you're trying to do a preserved deprecation strategy.
two more, while on treatment, it is appropriate.
It's been done.
We can get all the data before ..., I use it went up when possible. It may not be always informative. For example, if you have one of those Crohn's disease or subject colliders trial, when you have an induction phase on a maintenance phase, and went to trial, The patient completed that particular patient completed. All. The information for the induction phase was said, Yeah, go ahead and use it for the intercurrent events, happen.
The induction phase, So I don't see why you couldn't use that data. Now, the patient, what impacted by calling in 18 during the mainland phase? And I will say, yet, don't use that data, because you need to wait all the 48 week period in order to do quasi their response on our response.
Compulsively strategy is quite complicated.
I personally don't recommend for any intercooler around subtype discussed before to be included in the definition of the outcome.
The main reason is that it is almost impossible for example, to assume that all the patient impacted by co-ordinating should be considered as non responders or or assume that the station not impacted by Coburn 18, but that this quantity to you because they couldn't accident drug. They are also non responders. My recommendation, stay away from assuming that is a composite of strategy.
And it's better to go in it in a mining exercise for all day subcategories associated with it into the copying a team, and all the categories that we mentioned before, and doing evaluation one case on the time.
Let us see, the statistical consideration from the study, design perspective.
Planning style, the consideration for planning this, this study, I mentioned before, when we start working on protocol development, we make a substance showing off the dropout rate.
Effect size, variability, many other parameters for that, but certainly not only for the computation of the sample size, but we also make some shunning. When the interim analysis is going to happen, when the DMC's going to happen.
Goby 19, even when it's extremely hard to predict what is going to happen in the future, we're going to change. You need to be prepared in adding all this element.
So if you have it by trial and you're expecting that 35% coefficient of variation, and you know that you're going to have chef in the patient population that you're going to be using and where are you going to be a recruiter and many other factors that go into subject coefficient of variation most likely it's going to be higher.
But so this yesterday she tried to keep modeling very handily, do enough modeling to predict possible differences. React accordingly is the coefficient of variation goes up. And then ... adopted the sign.
Not because I am a subject matter expert in the field, but I do recommend when possible, try to define your study as possible about the design.
So you can include features or feature that will allow you to measure the total variability due to measure your effect size if you have to. And you can check for, for early termination, due to, to delete the boundaries whatsoever. That my recommendation at this point, recruitment rates, are gonna change, there are changes that is obvious. So essentially, with dropout rates and the interim analysis and BMC, I already mentioned the one you're stopping rules may change.
You shouldn't be predefined in all these adaptation, but if the protocol is written in an efficient way, you can even include the adaptation to go.
But, again, it will take a lot of negotiations with the regulators to ensure that the adaptation doesn't contradict the original plan, is not a major deviation from the original plan. Then how you're going to adjust for multiplicity and many other issues that may come with a new updation sample size, re estimation, or a plus.
strongly recommended.
My suggestion to this topic is, there are multiple attempts to try to do a blind samples .... Are assuming that you can do a blind, this sample size re estimation, uncheck the effect size. They find that somebody else can depend on version is due in the estimation. That's not the case. Remember, your online, the data and your X, you are checking for auditing beside the nuisance parameters. Then it is an unblinded sample surveys dimension.
If you don't have the option, or you don't want to start for efficacy, then you may be looking to expand in very tiny amount of Alpha. To justify the interim analysis, and they check that, you're going to be doing a year effect size, and then, you can keep the big proportion of your alpha to the final analysis. one of accommodation HTML attributes, for all employees.
This is a common question. People ask me all the time. Do you need to define esteem and only for their primary endpoint? The answer is known as the definition of the ...
and the nature of the signal is, to ensure that you have all the GPUs that will define the analysis and interpretation of each one of the endpoints. So, it is critical for the primary or the secondary, or tertiary or exploratory.
Even for a safety endpoint, Safety endpoint are as important as the efficacy endpoint's important. And it's any PK endpoint. Because, remember, again, co-ordinating may induce changes in the safety profile, because patient may be a positive, they don't know that they are because they are being tested. And they are Jerry's for the immune system somehow, has been compromised and they don't even know.
Missing strategies is also extremely important. Not only missing data strategies, for the end point of a cell, but also the missing data strategy associated with it. For example, quality of life instruments, or any other patient diary.
Missing components in inequality of light instrument, for example.
So all that definition to be defined an upfront corner reporting perspective. This is this is becoming a little bit of a challenge but it's important to have at least the following report data before, during and after the pandemic.
Many trials right now The they started recently generally to worry about before.
You didn't know when the pandemic is going to end but at least they're reporting during the pandemic. These gradients on electronic started last year before the pandemic was declared and you have a lawn Rush Study recommended to do it before and after.
Do sensitivity analysis in relationship to how many sites were close site that were temporary, close, or permanently code. Base it performed outside the window, is that if this window is representing major vertical deviation, then we need to discuss with our clinical teams to consider if that data will be consider an outlier, or you need to be done. Or what else to be done with that particular data. Changes in the protocol to Cesar.
And I feel that I mentioned before not only for the clinical and operational perspective, but also any change. There could have been, for example, if you are test, if you are reading heart rate with a particular instrument on site.
The high rate is not the measure in home, with the patient using a different device.
Then we need to ensure that the two devices are consistent with his shoulder. There is correlation of the measurement that is consistent between the measurement. There is reliability between the two measurement. Yes, use a heart rate, monitor, for example. But it could be any other one for an operational perspective.
The same is that a patient was supposed to visit this site, and I like using the Newsflash dystrophy type of trial where you need to take your kids to the physician.
And then, you're going to be doing the pull up this, or you're going to be doing stand-up desk whatsoever. Is not saying that if you do the same tests at home with their parent, than if you do the same test in front of the commission, the effect size and the response, you're going to be different because they are different.
The subject is exposed to two different scenarios premature.
They started to question, I should do, to, covering a team, going to make that count number of treatment interruption and whatnot also compliance with treatment as well. And then number of sub being impacted by calling a team. How many confirmed infection actually tested?
We can also have a possible, based on baser book, number of, based on the symptoms.
But they were not tested. So patients who experience going, 18, there were not tested. They saw that they have ..., they have all the symptoms, they have to be a way to validate if it is a critical we knit team. and not just through, but it is recommended to do both number of hospitalization due to common ID.
And of course, as I mentioned before, that, due to coffee and a team, where we measure the effect, where almost IBM, what we measure, the effect, is query, to distinguish between sensitivity analysis, type of sensitivity analysis, and then supplementary analysis.
Remember, the sensitivity analysis, the recommendation is change one, TPU, your primary endpoint the time. So if you're gonna be doing a sensitivity analysis with respect, the missing data assumption, and you're going to have multiple recently that assumption, you probably four ish intercurrent event category associated with common A team. And I am focuses and co-ordinating, but it could mean to do any other intercurrent events.
They tried to modify one into current event or a strategy to manage the equivalent event one of the time. Because if you do multiple at the time, it would be hard to perceive, to show where the P value is shifting or where the treatment effect is shifting or both. And whatsoever.
Remember, we can do sensitivity analysis due to the missing that assumption. And then we can do sensitivity Analysis with it in the Caribbean And in the corner of an astrology. But don't mix them both at the same time. You could but he's going to be more complicated.
Deep, important analysis are critical, very useful.
Not only using delta adjustment approach when you can either proportion of that respond and then do the shift, and then you can visualize the chest. I found very attractive use XRP scenario, especially for binary and poignant way. You assumed the missing data and proportionate whenever you responded. And I'll respond or the table sometimes end up being really big, but it is very informative. You just need to be mindful that if we have a lot of missing data than all, the combination of possible scenarios is going to be more complicated. I really love the idea of use and has tipping point analysis. If the WWC show not only of the shift from the P value, but also the chip in the treatment effect, and many other parameters associated with these clinical.
So, I recommend and enhance the point analysis, sensitivity analysis, also, respect compliant, with the protocol, that it will be assembled by using your, instead of using your full analysis, that you can use your, your per particle set. Or, if you want to use in your, modify for analysis set, then you can shift or switch the primary endpoint to use, and, therefore, the analysis. And the idea is to show compliance.
So, what is the difference between the patient population, where we can see they're complying with one or more parameters in the protocol?
And then, supplement or analysis analysis that will supplement.
You're not testing here, for, for sensitivity, to see when the P value cheerful, when the effect size changes.
Here, we're trying to find alternative method, right?
Using different covariates, different factor and visit different interaction effect associated with culminating in emulating how impactful doctor people are.
Data elements change the effect size.
subgroup analysis are important, but as we go with the group Analysis, my recommendation is always, remember, Everybody loves to C, P values, I recommend, instead, use in the 95% confidence interval or the SAML.
And if you're displaying P, values everywhere, remember to add a footnote, explained that those P values are ... in Nature, because sometime we have, on P values and the interpretation here, with so many possible.
possibility of so many possibility for doing analysis, based on copying a team of categories or cognitive issues. It can get complicated. Finally, implemented in the current events and ... strategies, different user couldn't really started even for the same integral around. Also for safety PK, PD employees, and I will be the one multiple times, Because there is a trend. I don't know if it's something that yes, me, that one of survey, that trend that we've seen. That means the corner, Randy's esteem and definition is only for efficacy, for every single employee in our clinical trials.
In general, if you can see, there is no one solution for all the evaluation of endpoints, or variables of interest in our clinical trials.
And then doing the analysis and consider as much as you can to include.
You can always e-mail me or e-mail our panelists today. Thank you so much for your attention. And I hope you enjoyed this presentation. Thank you.
Thank you so much for that.
Luis style is a really fascinating presentation and gives an insight into the effect that culvert 19 has had on the operational and statistical aspects of clinical trials. Like, I think there's so much attention right now being put on the vaccine trials. For, for example, ...
and Pfizer, and also for the therapeutic for COVID-19 Data platform trials you see for, like, the roadmap cap or recovery trials. But I think it's almost been forgotten in the wider community that there's all these other clinical trials, all these other conditions, which are trying to find new therapeutics for that. And that will face or are facing a huge difference. Or a fact of cover 19 on their ongoing trials. But also in terms of planning for trials in the World of COVID-19 until you know either a very effective vaccine happens or we come up with another societal strategy to deal with the COVID-19 pandemic.
So if people have any questions, there is a questions tab, on the right. We'll get to them after the webinar, either Luis, or myself, or get back at pending on the context of that off. The question itself, as mentioned in the slides Luis has given his e-mail for reference.
It is equal to Luis.rojas@parexel.com If you want to get in contact with us at NQuery.
It's an info@statsosl.com. I just want to ask you to take this opportunity to ask a few questions by watching presented here today, as, well as, talking about the first set of problems, which is, you know, you had an ongoing trial. And then culvert happened. It's almost like, you know, because, like many of these trials were, obviously, council for a period of time during the early stages when most countries underwent a very severe lockdown.
So, it's almost like you have a situation where you have to almost different trials that have happened, like different universes where people would have chosen previously. The more standard approaches kind of gotten all over the world. And then secondly, now they have one, which is being done virtually. And maybe it's only been able to only be able to continued in certain countries that have managed to get the disease under control. So, and then you're trying to stitch these almost together to maximize the evidence, because obviously sometimes you may need to remove that, get rid of some data because it's just not possible. But obviously anytime you have data, you feel bad about that. Because obviously other information on tying all of trailers and of subjects, which is not being used.
So like, I suppose like, from your perspective, though, from a practical and statistical point of view, like that process of stitching things together. How much has that affected here? Like, the engagements that you've had with the people vote as subject in terms of any, talks that you've had with people who've been dealing with subjects on the ground than the statisticians as well.
Yes, I was just kind of just kind of giving a general background there, but that was the main question I was just wondering, It's like you have had this basically no big event happen, which basically splits your trial if it was ongoing earlier this year and almost two parts.
You know, you have the pre COVID trial and you have a post cover trial. And now you have this big question of how you're going to stitch those together, but also, you know, how are you going to deal with the actual subject from the ground?
So in terms of like dealing with both the tri less than the statisticians and dealing with that problem, but also like the experiences people have had dealing with subject from the ground, like how much of a challenge has that been for, you know, everyone involved in the clinical trial community?
Yeah. This is an actual issue for many of us in the industry.
We just need to define, how are we going to handle the data, being in, in the overall data, because you cannot split it in half sometime, if there is enough time, and you're seeing a major deviation from the protocol.
For example, of, when you have a therapy, that is like a single infusion or a single therapy, and if the patient doesn't have oxygen to the particular therapy, then you can do a sample side replacement. For example. In some cases, when you have multiple drug administration, and the less than 80% compliance, for example, is not possible, then you see a dramatic change.
Sometime, you can see that many trials are being put on hold. Some of that trial had been completely stop.
Because at the end of the day, the submission will take into account the original patient population and the original plan that was set up into into trials are easy. Especially when they try when I say that. Because I know that the design, then it will be very little room and not other thing to do. Yes, to shut down the trial.
I hope that I answered your question.
Yeah, that's, that's a great answer. And obviously, a lot of the strategies here are very useful in terms of dealing with that. But I suppose, you know, when you're designing a trial, right now, in terms of the practicalities of, how much would you recommend, like, how strong do you think these tools are to deal with the issues of, like, covered like spikes happening in certain countries. Or how much should you be thinking aim to almost I don't want to say Work around Covid-19. But to perhaps focus resources at the moment on like regions or areas where you believe that they're more likely to have a stable covid 19 profile. For example, we know there's a fairly differential responses in far east and countries like, China, South Korea, versus in Europe and the United States.
Actually, I am very proud of working on Parexel and one of the thing that I really enjoy here is we've been extremely proactive in identifying the possible impact of ... head of the game. So, we know, for example, that lease winter season here in the US. These cases are going up, so we have something that we call a virtual clinical trial design studies.
And then we have, for example, with a group, team members, ensuring that we have the resources that the infrastructure and the technology to switch to virtual visit, or even started to travel with virtual visit. Or do that innovative approaches to ensure that we don't lose that data that we keep on track.
So we continually monitor site regions. And then if there is a need to open a new region, then we co-ordinate on each other to ensure the reaching the Marlins, especially in the trial as current events move up and down the multiple regional underground.
You know, the classic statistician problem of these things coming in too late.
No, You cannot ignore the problem you need to look at as actively and deal with strategies. And another example that I faced recently is, usually, when you assume, well, in this type of trial, we're spread like a 10% dropout.
You need to triple the number in the copyright theme, in the common item era. Because something that I've seen is, do not be over optimistic that you can keep up with your recruitment rates.
Keep an eye in how the number of positive cases are going up in multiple regions, and plan around those issues.
Check, mean daily basis, what radios are getting shut down, in which region, how you burn restriction, because that will happen everywhere. And something that I mention about the ... is also critical, and I can give you my personal opinion on my personal experience. Yesterday, I had an appointment with, my cardiologist, and this I didn't know are now gone, and I don't want to go for any of these risks.
Ah, I will call him, doctor that everything inside, and at the end, I'm going to wait a little bit longer, and then you can see that there is not an expectation to the visit.
Is the risk of getting expos, to culminating in the moment, where in my region is going, fast way to it is going way too fast.
Wasn't really that the main topic today, but I didn’t want it stealing the light, and many of these other presentations. But before we finish up, I don't know, you have any strong opinions on any of the current trials for covid-19, either vaccines or therapeutics, like some of the innovations we've seen. There are some of the changes there.
Do you have any comments on that before we finish up?
Yeah, I gotta be very careful.

Even when I understand that, we're trying to do to give, some benefit. We're trying to push on peace of mind in our patients. The same time, we need to remember ourselves that we can not induce bias in our technical problems. Are they doing what they are supposed to be doing? Yes, to a certain extent? It is necessary to encourage patients volunteer to participate in clinical trials. Yes, by the way, it is being used, especially by politicians. In my humble opinion is not an optimal Peter, why we need to encourage patients to participate in clinical trials, BTC Yeah.
But we shouldn't be provided. Any information about the possible resolve of a vaccine or or any other retreat before covid-19 people. Died. Induced. Again, like I mentioned before, the Civil Effect patient may have way too much expectation that is going to work from a positive perspective, or they may be the opposite. You say, No.
And he happened to hear the USA when people say, No, I don't want to get tested, which I don't recommend. I feel that we allow me to participate in the control. We need to support all those screen controller, But, again, the way it is getting communicated, I've seen that, can use an improvement.
I definitely agree with that point, and it's a very important point to make. I think, you know, it's hard.
truth is always preferable to to an easy, you know, half truths, you know, and I think we have cautionary tales from the recent past.
For example, take tamiflu was one of those cases where, you know, a lot of urgency sometimes creates some perverse incentives that we need to be aware of, And as statisticians and scientists, it's important for us to go out there and communicate in presentations, such as this, Obviously to other practitioners, but also preferably to other people out in the real world community as well. Like, it's always quite, well, you know, I don't want to have had the most success talk about statistics, what people like my family, for, example, of, what I I didn't make sure to try. And if I make it back for Christmas, I will share. And they kept trying again and again.
But, yeah, like, when, there's so much money, and so much expectation, that cost, we do need to be very cautious, thereby not to allow the lower half. True. To get far ahead of where we actually are, like, vaccine development is incredibly difficult, Therapeutic development is very difficult, and we need to be temporary expectations a little bit, despite the second.
I think that is our, we're mandated to do, to enforce integrity in our clinical trials. How do we could even, nobody can even think of punishing one of us increasing our opinions because that's what I would call the ugly through. The obstacle is very well defined in ISO in GCP to resolve the clinical trial is now go in under any circumstances. So copywriting is not an exception.
Being that we must observe the respect for our rules of engagement, a trial that will ensure more transparency in the results. People will trust more our result if we keep on doing what we're doing, which is compliance compliance, and more compliant with ... and GCP.
Yeah, and it's, like, this is, as you say, it's a long run game.
Like it take, it took decades of hard work to create the current infrastructure, which allows people to trust that their therapeutics and medicines are safe and are effective. And they would be a shame to undermine that. Because obviously, in terms of, in terms of when the public cares about that stuff, it's only in a very few circumstances, such as the current ....
This obviously affects every single approval that will pull exist going today forward.
So I think that, I think that probably gets us up to the finish of this.
So, like, once again, I want to thank you so much, Luis, for taking the time to do this webinar today. As mentioned, if you have any questions or questions out there, we'll get to them afterwards or you can e-mail Luis directly and females in the slides. Or you can e-mail us at nQuery at info at ... dot com.
We will be sending a recording of this afterwards, so don't worry about that. I believe the slides should be sent alongside that. So at any further ado, once again, thank you so much, Denise, and thank you so much for everyone for attending, and I hope to talk to you soon at our next webinar.
Thank you, everybody.

Subscribe by Email