RLS Protect logo white

WEBINAR

Trial Disclosure: A Focus On Rare Diseases

Speakers: 

Elliot Zimmerman, CEO Real LIfe Sciences
Ahmed Eladafrawy, Implementation Specialist & Health Canade Liason, Real Life Sciences

Disclosure, Transparency & Challenging Populations

Working with rare disease populations requires compliance strategies be thought through strategically in advance. Small populations, like those found in rare and ultra-rare disease populations can increase the likelihood of patient re-identification if advanced methods are not applied. This webinar session will focus on learnings and best practices to apply when working with rare disease trials and populations.

Meet Regulatory Demand

Cutting Edge Data Anonymization &
High Speed Redaction.

Are you interested in staying ahead of the increased Anonymization requirements for Health Canada PRCI & EMA? Do you want to learn more about new regulations or the changing expectations from Regulatory Bodies? We are committed to informing the Clinical Trial Disclosure & Transparency community. We educate & partner to advance participant privacy and data sharing. Please reach out with any questions you may have or to schedule a brief 15 or 30 minute chat to discuss moving you forward faster.

    * required field

    Webinar Transcript - Trial Disclosure A Focus on Rare Disease

    Alex Caracappa:

    We have an exciting webinar for you today entitled Trial Disclosure: A focus on Rare Diseases. We have two speakers with us today, Ahmed Eldafrawy, disclosure and risk specialist, Health Canada liaison at Real Life Sciences and CEO of Real Life Sciences, Elliot Zimmerman. With that, I'll turn it over to Elliot to get us started.

    Elliot Zimmerman:

    Thanks, Alex. Appreciate you getting us kicked off today and thank you to our audience for taking a few minutes out of your schedules to join us today.

    Elliot Zimmerman:

    As Alex mentioned, we have a full agenda for you today focused on rare disease and disclosure best practices. What I wanted to start out with though is a little bit about our vision at Real Life Sciences in particular. Our business is focused on using technology and natural language processing to protect patient privacy.

    Elliot Zimmerman:

    Specifically Real Life Science has two solutions that we offer to the life sciences industry. Real Life Sciences Reveal, which is a natural language processing based solution focused on leveraging social media real world data, to understand the impact of quality of life in different diseases.

    Elliot Zimmerman:

    Real Life Sciences Protect is the solution that focuses on disclosure in particular, in support of various regulations and voluntary disclosures that some of your organizations are probably participating in today as well.

    Elliot Zimmerman:

    Based on our experience with the disclosure arena, we wanted to share some of our learnings and experiences with it as it relates to rare diseases. In particular, we've done numerous submissions for disclosure purposes through EMA policy 70, Health Canada PRCI, I'm sure these are all familiar, as well as supporting various pharma companies with their voluntary sharing initiatives as well.

    Elliot Zimmerman:

    We process data sets using a quantitative methodology and of course we process documents for anonymization purposes including redaction, anonymization, and transformations. We've done a lot of this including rare diseases and so we took away some of our learnings and organized that into today's content to share with you some of the things that we've learned and of course would like to hear from you on your experiences as well as it relates to rare diseases and some of the challenges as it relates to disclosure purposes.

    Elliot Zimmerman:

    Our agenda today for the first few minutes I'll speak in general about rare diseases just to get everybody oriented and there might be some refresher information for some of you, but I think also some interesting facts to take away about rare disease in particular. Then we'll focus on the next level of detail around what are these different implications and what does all this mean for purposes of disclosure when you're dealing with small patient populations? Ahmeds going to take us through some of the specifics and some of the tactics around how you can tackle this and make sure you're solving this in a smart way.

    Elliot Zimmerman:

    Let's get started with a little bit about rare diseases and there's some interesting facts on this slide and the next one. Some of these things you may already know, but it'll act as a good refresher to get everybody oriented to today's topic.

    Elliot Zimmerman:

    First of all, the FDA does refer to any rare disease is a disease that impacts 200,000 people or less in the United States. Of course, there's also the ultra rare category, which is just 6,000 people affected or less. With this in mind, don't forget that the Orphan Drug Act was passed in the 1980s to encourage more research and development and innovation in supporting these rare diseases, but as you'll see, there's still quite a bit of work to do in this particular area. It's an ongoing problem and an ongoing investment area for a lot of pharma organizations.

    Elliot Zimmerman:

    Here's some other interesting facts and I think you'll probably learn a few things that are kind of interesting takeaways from this as well. Just in terms of the number of rare diseases that impact people around the world being over 7,000. When you look at each one of these individually, any particular rare disease, the numbers are indeed small, but when you look at them in aggregate, there's a huge number of people in the United States. Of course in Europe and globally that are affected by rare disease.

    Elliot Zimmerman:

    Sometimes an individuals even impacted by more than one rare disease. The research indicates as well that it takes a long time to get properly diagnosed. In a lot of cases, 95% of these rare diseases have no treatment. It's a very significant problem. When you think about quality of life and other impacts of these rare diseases.

    Elliot Zimmerman:

    Here are just a few examples and the number of people that are impacted by some of these and some of these are very, very small with just a thousand people or 285 people. In other cases, more on that upper threshold of what qualifies as a rare disease like CREST syndrome of just under 200,000 people. Just some examples, and obviously the list goes on and on.

    Elliot Zimmerman:

    That tells you a little bit about the general problem around rare diseases and in particular, what we wanted to call out next is how does this really translate to transparency and disclosure initiatives? What does it really mean when you're dealing with disclosing a study, or a trial, or set of trials focused on a rare disease and how to disclose the results without jeopardizing patient privacy?

    Elliot Zimmerman:

    Of course the same kinds of challenges exist in broader studies as it relates to protecting company confidential information. Of course, however, what we're really focused on here is around patient identity, protecting the patients, and how to do that when you're dealing with small patient populations. Of course also the disclosure requirements are no different for a rare disease as they are for a non rare disease.

    Elliot Zimmerman:

    From a regulatory perspective, as well as your internal sharing policies, you need to adhere to the same kind of measures and the same expectations in terms of what you're disclosing and how you're going about it. That said, there needs to be more attention and acute focus on how to protect that patient identity. We have ideas and experiences that we want to share with you and from what's worked well for us with previous submissions with Health Canada and in the EMA as well.

    Elliot Zimmerman:

    All that said, one of the takeaways you'll find from today's session is the importance of considering a quantitative methodology to leverage the data sets from your trials, to empirically evaluate the data, and have that help drive decisions around how you disclose the documents. Working in concert together how that can help to overall make sure that you're, de-risking your disclosure initiative overall.

    Elliot Zimmerman:

    With all that said, we'd like to initiate the first of two audience polls. This first one, and you should see it up on your screen now, is just asking the simple question around what level of experience does your organization have in working with rare diseases in particular around disclosure projects? Maybe you're just getting started and for others, this may be your business focus in particular so you already have some, some experience and we'd like to learn a little bit more about our audience today and what that looks like from an experience perspective.

    Alex Caracappa:

    Yep. It looks like it's coming through here. It looks like mostly some experience and the second most popular one is limited or no experience.

    Elliot Zimmerman:

    Okay, great.

    Elliot Zimmerman:

    Let me just close this window on my screen. Perfect.

    Elliot Zimmerman:

    Thanks for the participation. We'll have one more of those polls a little bit later on and they're really helpful just to get some perspective on how you all feel about these topics as well.

    Elliot Zimmerman:

    Let's get kind of into the next level of detail around what are some of the specific disclosure considerations and the challenges that we need as it relates to these rare disease populations and getting into disclosure requirements in particular.

    Elliot Zimmerman:

    First, we wanted to start with the regulatory guidance and in particular we have experience with all of these regulatory bodies as I'm sure that you do as well, but we pulled some excerpts from, for example, more recently, the CTR and the EMA website. Just in the last few weeks I actually think it was the first week of April, the EMA published guidance around anonymization in support of the CTR using the CTIS.

    Elliot Zimmerman:

    There's some specific comments that the EMA published around rare diseases and using a quantitative methodology. I won't read through those specifics for you, but we did include the links to this material for the CTR related comments from the EMA, Health Canada PRCI, and the regulatory bodies guidance in general, as well as some commentary around rare diseases. Then also around policy 70 from the EMA.

    Elliot Zimmerman:

    Those three links are live in this particular slide. If you're interested in this to be able to find more information around it, just contact Alex, we'll share his email address again at the end of the webinar and he can email you these slides so that you can leverage these links.

    Elliot Zimmerman:

    In terms of this quantitative methodology approach, what this is really doing is leveraging more of a statistical approach to evaluating the data sets from the trials that are applicable for your submission. That allows you to evaluate empirically the data and the patient population characteristics in order to make decisions in a more statistical manner around what should be retained, what information should be transformed, and what may need to be redacted entirely.

    Elliot Zimmerman:

    What we've done here is a very straightforward example where we have qualitative on the left and a quantitative approach on the right just so that you can see visibly some of the differences. First of all, a qualitative methodology is generally going to be more rules based. It also may be a set of rules that are applied across studies and across trials and submissions and therefore they're a little bit more generic.

    Elliot Zimmerman:

    What we find is, especially for rare diseases, you really want to look at it on a trial specific basis in terms of making these decisions around what to retain, what to transform, what types of for example, how many years in your age bands may you want to use? In this particular example, a 20 year age band on the right side from a quantitative approach is what was derived and decided based on the statistical analysis.

    Elliot Zimmerman:

    At the bottom of the quantitative approach you'll see just an output. This is from our particular system. What it's showing is the different scenarios that were analyzed systematically, which one ultimately the system recommended in terms of coming up with the particular scenario around what will allow us to balance risk of re-identification with maximizing the data utility for this particular trial.

    Elliot Zimmerman:

    This of course can be applied for a rare disease trial as it could for a non rare disease trial, but in this particular case and in this example, the risk threshold was set to 5% in particular because this is a rare disease. In a non rare disease you may use a higher threshold such as 9%, for example. The system takes that into account and makes the recommendations for the user.

    Elliot Zimmerman:

    With all that said in kind of setting those particular examples and comparing qualitative and quantitative, I'm going to hand it off to Ahmed now and he's going to go through six or seven specific examples and tactics around managing this process for rare diseases.

    Ahmed Eldafrawy:

    Awesome. Thanks, Elliot. Appreciate that. Yep.

    Ahmed Eldafrawy:

    Here we have laid out an overview of some of the key rare disease considerations that we have felt have enabled us to have successful rare disease submissions in the past. What we'll do is go through each of these and provide some insight on how each of these considerations can be part of your rare disease disclosure.

    Ahmed Eldafrawy:

    The items that we're going to go in depth are going to be as you can see on the left column maximize data utility through use of similar trials, assessment of risk of re-identification for small patient populations, quantitative results analysis, and also providing that regulator and internal justification and gaining those approvals, identification of outliers, redaction of aggregate level data, transforming at the category level as opposed to full redaction, and analyzing risk against attacker models.

    Ahmed Eldafrawy:

    Ultimately, we felt that taken into account each of these for your disclosure help you meet your data utility criteria set by regulatory while still protecting your patient's data. We've also included the anonymization approach on the right side of your screen to kind of show you how each of these considerations can be done.

    Ahmed Eldafrawy:

    As you can see a quantitative approach can help achieve all of these considerations while a qualitative approach can help with a couple. In the next few slides, we're going to kind of take a deep dive into each of those and show how you can factor those into your rare disease disclosure.

    Ahmed Eldafrawy:

    The first one as we mentioned in the previous slide is reference population. As we've mentioned earlier population sizes for rare disease studies are typically very small. Performing a risk assessment just on the pool of the subjects you have ultimately lead to high redaction because you just don't have enough subjects in your study or in your data sets to make equivalent classes that are large enough to retain or anonymize certain attributes within your study.

    Ahmed Eldafrawy:

    So because of that, one thing that we recommend you do is using a database like clinicaltrials.gov to try to find similar studies. Once we find that pool of studies, we can filter their characteristics to make sure they coincide or overlap with our study. A few of the attributes that we would filter for to make sure that they overlap with our study would be age, country, gender, and dates. Once we filtered those similar studies, we can then extract that population and use it and supplement for our reference population within the risk assessment.

    Ahmed Eldafrawy:

    The next consideration that we are suggesting would be, of course, then assessing your risk of re-identification. Obviously from a quantitative approach, that's going to be something that can be set and defined. It's important consideration specifically here because you're dealing with a highly sensitive population in a rare disease study. You need to make sure that the risk threshold is properly defined and met.

    Ahmed Eldafrawy:

    We define the risk threshold of typically we do it by using a privacy model like K-anonymity and then we can define a value for that privacy model that then dictates the exact threshold we want our outputs to meet. The higher the value we for the privacy model, the lower the risk threshold will be.

    Ahmed Eldafrawy:

    As you can see on the image here, the higher the K value, it means that the equivalent classes or your equivalent groupings are going to be larger, which then makes each of your subjects characteristics, or attributes less identifiable, therefore it lowers the risk.

    Ahmed Eldafrawy:

    In addition to this K-anonymity model, we also use a K-map model and supplement to that and that model specifically factors in your reference population that we use for the risk assessment. Kind of using this K-map intangible with the K-anonymity model for reference population it gives us an opportunity to preserve some more data utility we would have otherwise lost while still meeting the risk threshold.

    Ahmed Eldafrawy:

    Here just kind of from evaluating quantitative risk scenarios, the main thing we want to portray is that with quantitative risk, not only do we just ensure that we meet the risk threshold, but we know that quantitative risk assessments attempts to preserve transformation options with high data utility. It provides you with multiple transformation options for the user to evaluate.

    Ahmed Eldafrawy:

    Having the opportunity to evaluate those different scenarios is important with rare disease studies specifically because oftentimes we know that again, it's a very limited and small population. We know that sometimes we can't retain all the attributes we want due to the nature of the study. Having multiple scenarios often can give you an opportunity to almost handpick one or two attributes that are very clinically relevant that you would want to retain for disclosure.

    Ahmed Eldafrawy:

    Then we can evaluate these options that you can see here on the left and see if that's feasible. Just kind of to take a deeper dive into what that screen that you see on the left is. Again, it's a screenshot from our application and it's showing you the different transformations that have been generated for one risk assessment.

    Ahmed Eldafrawy:

    As you can see, each transformation is represented by one row and within each of these rows it's providing you a set of numbers. Each number within every row represents a quasi identifier and then that quasi identifier has a value. Then you can assess what the anonymization approach has been based on that value.

    Ahmed Eldafrawy:

    For example as I mentioned we may want to retain a specific attribute. So we can evaluate all these different transformation options and then see that attribute that we're trying to handpick or look for maybe let's see which of these options has it set as it's value of zero because we know a zero representing a quasi identifier in this case would mean that this value can be retained. That's one of the benefits of kind of having this ability to be able to do that.

    Ahmed Eldafrawy:

    In addition, just one final thing having this option that we can generate multiple scenarios that meet the risk of re-identification is also helpful from a regulator standpoint as well.

    Ahmed Eldafrawy:

    For example, what we would do in this situation is during a PIM or meeting with a regulator, we can often present them with our transformation option we selected, but also show them that we've evaluated multiple scenarios to come up with this conclusion. To show that we've done our due diligence in selecting not just any option, but we've selected the best option, preserving high data utility, and also being clinically relevant as well.

    Ahmed Eldafrawy:

    Next thing we wanted to take a look at is going to be identification of outliers. In this regards outliers within a quantitative risk typically indicate that there will be records that will have their indirect identifiers or their quasi identifiers suppressed. The reason for that is because the risk assessment may be deeming them as high risk subjects.

    Ahmed Eldafrawy:

    As you can see and basically they cannot be grouped in equivalence classes that meet that risk threshold. There is a possibility that you can retain them if you need to, but then the flip side of that is you're going to compromise the data utility for the rest of your subject population.

    Ahmed Eldafrawy:

    Just to kind of give a brief example of that is if we have a rare disease study with 65 subjects and we have four subjects that we've considered outliers, we're going to suppress the quasi identifiers for those four subjects in order to be able to retain an attribute maybe like weight or BMI for the rest of the subjects.

    Ahmed Eldafrawy:

    If we want to retain the quasi identifiers for all of these subjects, then we may have to then just redact that one attribute for the whole pool of subjects that we have in our study. That's kind of the things that we typically evaluate in quantitative risk analysis.

    Ahmed Eldafrawy:

    Just most importantly here we've successfully initiated this approach of including suppressed records to preserve the data utility for the greater whole of the subject population and regulatory authorities are typically in support of that because it does provide that additional data utility for disclosure.

    Ahmed Eldafrawy:

    Here in regards to aggregate data typically we don't recommend redacting aggregate level data since it's already a form of anonymization, but with rare disease studies, we sometimes need to take some additional measures. We may typically redact it if it's going to be rare disease terms, highly sensitive terms, or if it contains low end values. Ultimately again, just we let the quantitative analysis on the patient level dictate that for us then we can make that evaluation if redaction is necessary or not.

    Ahmed Eldafrawy:

    Another consideration we have is going to be transforming at the category level. This such scenario is specifically important for when full reduction is required. Again, the idea here is we're attempting to preserve the data utility, but as we've mentioned earlier in the presentation rare disease studies are not exempt from transparency expectations. You still kind of have to make every effort to make sure that you're still doing that, even though it comes with its challenges.

    Ahmed Eldafrawy:

    One thing that we can do that has the same impact as with action, but with much better data utility is categorical transformation. Categorical transformation it's going to provide better data utility because it replaces the original name or value of a term with the name of the category as opposed to just full redaction or as the black box.

    Ahmed Eldafrawy:

    On this slide here, you can see we have a page within a document that requires based on our risk assessment, it deemed that these three subjects require full redaction for all of their attributes. Here's kind of two versions on how we can do that.

    Ahmed Eldafrawy:

    On the left, you have the version with the full redaction, which you can see, we just basically black boxed all the attributes and on the right, what we did basically replacing the attribute with the name of it categorically transforming the attribute. Obviously, disclosing a version that we have on the right looks much better and provides much better data utility and clinical utility for disclosure.

    Ahmed Eldafrawy:

    Finally, the last thing we like to consider for quantitative risk analysis for rare disease disclosures is analyzing the risk. We want to be able to evaluate the risk thresholds based on different attacker models. We know that at this point would be any transformation option we've selected meets the risk of re-identification, but we just kind of, again, for rare disease considerations, we want to take it a step further and make sure to see how it measures up against different adversary models.

    Ahmed Eldafrawy:

    What we do is basically just provide some additional reassurance to ensure how we've successfully anonymized the data. A few of the adversary models that a quantitative tool can evaluate are going to be the journalist risk, marketer risk, and prosecutor risk.

    Ahmed Eldafrawy:

    In regards to rare disease studies, the main one that we pay special attention to because it's also the most conservative approach out of the three will be the prosecutor risk. Then obviously showing and then kind of showing those additional risk metrics and providing that to the regulators kind of buffers your approach, and kind of your stance on the approach that you're suggesting. More specifically the transformation option that you've selected.

    Ahmed Eldafrawy:

    This is connected to the previous slide. This is just a couple screenshots kind of analyzing the risk of those three adversary models we've just discussed. On the left, you can see it's our input data, and you can see that the risk of re-identification across those three adversary models is much higher. As opposed to on the right, which is after anonymization the risk of re-identification that's measured up against these models is much lower.

    Ahmed Eldafrawy:

    Just kind of having that additional reinsurance really helps us kind of gain traction with our regulators in support with our anonymization approach that it's safe, providing high data utility, and you're still disclosing quality clinically relevant information.

    Ahmed Eldafrawy:

    Finally, we just have our second poll question. My organization has experience with quantitative risk methodology, extensive experience, some experience, or limited, or no experience?

    Alex Caracappa:

    Looks like answers are coming through here. I'll give it another second.

    Alex Caracappa:

    Looks like limited to no experience is the most popular answer followed by some experience.

    Elliot Zimmerman:

    All right. Thank you, Ahmed and to wrap up this portion of today's webinar I wanted just to share with the audience that we've kind of summarized in this particular table for purposes of being a good reference that you can take away if you'd like these slides. We've listed out all of those different components and attributes that Ahmed was walking you through with just a summary of kind of what challenge that represents in particular for rare diseases and what potential solutions you can apply.

    Elliot Zimmerman:

    I think in those examples that Ahmed was sharing, obviously you saw a few screenshots of the RLS solution, but if you're doing this on your own, or have another provider, or whatever the case may be, these concepts still apply. These are still the same challenges that you're going to be dealing with from a rare disease perspective and of course, we're happy to help you as needed.

    Elliot Zimmerman:

    With that in mind from a Real Life Sciences perspective, we build this technology that you heard on Ahmed speaking about and we use it on a day to day basis to support our customers as well. This is specifically designed for the life sciences industry and support of these types of disclosure requirements. We do have an outsource managed services team as well using our solution on a day to day basis and helping with these submissions.

    Elliot Zimmerman:

    Lastly, of course, we help a lot of our customers as well with just their overall disclosure and planning process around disclosure in support of regulations and their voluntary policies as well to make sure that they're getting their biggest bang for the buck in terms of consistency of the approach and the methodology that they're using so that the results can be reused across the different regulations and for voluntary sharing purposes as well.

    Elliot Zimmerman:

    With that, that completes the presentation portion of today's webinar. Alex, I'm going to hand it over to you to help facilitate questions from the audience.

    Alex Caracappa:

    All right thank you, Elliot.

    Alex Caracappa:

    Looks like we do have a few questions that were submitted in the Q&A. If you do have any questions and haven't submitted yet, you can use the Q&A at the bottom and put them in right there.

    Alex Caracappa:

    It looks like the first question that we have here is, "What do you recommend if similar studies are not available to obtain a reference population?"

    Ahmed Eldafrawy:

    Yep. Hey, thanks, Alex. I could probably take that one.

    Ahmed Eldafrawy:

    Again, with rare disease studies that can be a challenge. Although for getting a reference population ultimately similar studies is our preferred approach, but there are additional considerations you can take.

    Ahmed Eldafrawy:

    One can be looking for patients in trials set by a specific study sponsor. Another approach can be to supplement that reference population will be all patients in a specific geographic area as well. That can based on where the study was conducted. These are a few other alternatives to provide you with a reference population.

    Alex Caracappa:

    Okay, great. Thanks. We have another question here. It says, "How does K value directly impact the size of the equivalent classes use for anonymization and how does it affect the risk of re-identification?"

    Ahmed Eldafrawy:

    Again, I can take that one. That one, it kind of goes back to the anonymization approach with K-anonymity. The K value that we define is going to be directly tied to the equivalent classes that basically the algorithm within a quantitative tool is trying to accomplish.

    Ahmed Eldafrawy:

    Really here, the K value is going to indicate that an equivalence class with that number must be met in order to meet the risk of re-identification to retain or anonymize that attribute.

    Ahmed Eldafrawy:

    Really the key element here is the tool is trying to make equivalent classes to see if the value can be retained or anonymized. Otherwise, that's where we go into full redaction if your K value does not equate to equivalent classes that's large enough to do so.

    Alex Caracappa:

    Okay. Thank you. We have another question here says, "What helpful resources can we review to learn more?"

    Elliot Zimmerman:

    I'll be happy to take that one, Alex.

    Elliot Zimmerman:

    Resources on I assume the question is around rare disease disclosure. I think a couple recommendations. First of all, in that one slide I referred to earlier, we've got the hyperlinks to some of the materials from the various regulators. That's a good resource.

    Elliot Zimmerman:

    On the Real Life Sciences website we have I think at least one case study specific to rare disease disclosure that you could take a look at to learn a little bit more. What else?

    Elliot Zimmerman:

    Yeah, I think those would be good places to start. Of course, if there's other specific questions, reach out to Alex and we're happy to set a call or converse through email to help address any other particular questions that you might have, but those would be good resources I think to start with.

    Alex Caracappa:

    Okay, great. The next question we have here, it says, "You mentioned using 0.05 or the 5% in Health Canada PRCI submissions. As Health Canada can be quite wedded to 9%, even in very sensitive disease areas. Do you have any pushback on the more stringent risk threshold? And if so, excuse me, how did you convince them to accept the 5%?"

    Ahmed Eldafrawy:

    Thanks Alex I can probably take that one as well.

    Ahmed Eldafrawy:

    We have been able to produce a risk assessment with a 5% risk threshold for rare diseases in the past. It's not to say that it always have to be 5%. We would preferably recommend 5% for rare disease studies, but ultimately it kind of comes back to providing that justification, which can be very helpful.

    Ahmed Eldafrawy:

    When we analyze the risk and we are preferring or suggesting the specific approach being able to kind of, again, buffer our approach with the metrics to say, well, this is kind of what the risk of re-identification is with 9%, as opposed to 5% based on the context of this specific study.

    Ahmed Eldafrawy:

    Again, that's going to be the key element here with, with quantitative analysis is the context of this drug and study, the amount of subjects you have, the data that's present and the attributes that are available. All that is being taken into consideration. It's never one specific or one only way to do it. We really kind of have to factor in the scenario and the challenge at hand.

    Alex Caracappa:

    Okay, great. Thank you.

    Alex Caracappa:

    Looks like we have a few more questions here. Next one is, "What is the recommended suppression limit value?"

    Ahmed Eldafrawy:

    Yep. Again, that's going to be one of those questions that can be based on a per scenario, but ultimately when we've approached it with the regulators, we've been able to have a suppression limit of up to 20%. It's not necessarily saying that when we input a value of 20%, it means that the quantitative analysis always suppresses 20% of our subject population.

    Ahmed Eldafrawy:

    It just means that we're providing it with ability to go to suppress up to 20% of quasi identifiers. Sometimes that can also be a value that we define with our customers to make sure that it's something maybe somewhere in the middle so we're not overly suppressing that either.

    Alex Caracappa:

    Thank you. The next one here, let me see, "How does quantitative risk dictate which attributes should be categorically transformed?"

    Ahmed Eldafrawy:

    Oh, sorry. I was on mute. I can probably take this one as well.

    Ahmed Eldafrawy:

    Again, so with the quantitative risk, what it will do is based on the transformation options that you've selected, or excuse me, the transformation options that get generated, again, you're going to evaluate and then decide on a transformation rule that you see is most fit.

    Ahmed Eldafrawy:

    In that scenario for one of the attributes based on that rule set, you may have to redact the specific attribute. We then kind of take that information and from the redaction output of that we can then categorically transform it. That's going to be one approach.

    Ahmed Eldafrawy:

    The other approach is going to be for our outliers where we know that they need full redaction of all their quasi identifiers and then we can categorically transform all the attributes for those suppressed subjects.

    Alex Caracappa:

    All right. It looks like we have one more question here, "How many rare disease type disclosure regulatory submissions has RLS completed?"

    Elliot Zimmerman:

    I'll take that one, Alex give Ahmed a little bit of a break. How many submissions for rare disease have we completed? I think we've done at least four or five at this point and we have another two in process right now.

    Alex Caracappa:

    All right, great. It looks like that was our last question. If anybody else does have any other questions you can feel free to email them over to me and we'll get you an answer, but we'll move on to let you know about our upcoming webinars.

    Alex Caracappa:

    Our next webinar will be on Thursday, June 23rd, called Disclosure Best Practices. It'll be focused on commercial confidential information, CCI for EMA submissions, and confidential business information, CBI for Health Canada submissions.

    Alex Caracappa:

    If you'd like to sign up for that webinar, you can email me or better yet. At the bottom of the screen, you can go to rlsciences.com/events. As a reminder, if you'd like a copy of these slides or a copy of the recording, please feel free to email me and my emails right at the bottom of the screen here also.

     

    crossmenu