Matinum

Taking Charge of Your Health


Thank you Eugene. Good morning, afternoon
or evening, depending on where you are in the world with us today. Thank you for joining
us on this webinar. There are a lot of slides to get through today, so I apologize for the
brevity in some of the slides. But, as Eugene stated, we will be delivering both the recording
and PDF copies of all of this information and the academic references to you, um, as
part of participating in this webinar. So again, thank you. So, today, what I wanted to cover was a brief
overview in the common language of pharmacoepidemiology and some of the different observational study
types that are utilized in signal confirmation and signal, sometimes signal detection as
well. I also wanted to focus specifically on two
very interesting FREE online databases and tools that you could use in your pharmicoepidemiologic
endeavors. And then I will also cover, very quickly, the traditional fee-based databases
that are used in these types of observational trials. So, first, if we just look at some of the
common language, just a few small acronyms: ADR: Adverse Drug Reaction. I also added on
here a PR for product reaction We do a lot of work with medical device firms,
and the construct of ADR is not always appropriate for a device firm. So, I have changed that
to an acronym of APR for common discourse. Also, the CIOMS working group. These will
be, these are the various meeting teams that are the Council for International Organizations
of Medical Sciences, who routinely meet and cover very important pharmacovigilance and
drug safety strategies for the work that we do in pharmacoepidemiology. And these slides
are very loosely based on the CIOMS 8 working group specifically. And then, of course, we have the EMA, the
European Medicines Agency. We also need to come to a common understanding
of what we mean by the term signal through these slides. And I took this from the CIOMS
working group application. If you have not received a copy of that or reviewed it, I
highly recommend it. It is a really good, a very comprehensive read when it comes to
signal management in both traditional and non-traditional means of accomplishing your
company’s goal of doing signal detection and the other parts of the signal management life-cycle:
Signal Prioritization and Evaluation. Through these slides, we want to see a signal
as information that may arise from one or many other sources that suggest some type
of new causal association or a new affect of a known association. So, for example, an
increase in intensity of a particular association or maybe even a decrease in intensity of a
known association. Signal detection itself is just simply the
act of looking for these various signals and the CIOMS 8 working group publication is very
clear that while the use of comprehensive and large public database sets, are really
interesting and cutting edge right now, that the traditional individual case safety reports
and the periodic report assessments that occur from the individual case safety reports really
should not be overshadowed by these interesting, automated signal detection methods such as
Empirica Signal. In traditional signal detection is just as important as the automated signal
detection method that you may be familiar with. I took a definition from a British researcher,
Waller, who published a text that is also in a slide deck in the reference section.
A very easy read, a very comprehensive introduction to pharmacoepidemiology. Again, I highly recommend
it. In his text, Waller describes signal prioritization as really a controversial method of ensuring
that only those signals that are deemed worthy of internal resources are passed into the
formal, more complex and comprehensive evaluation process. And, I submit to hear that internal
resources does not just indicate a monetary concern. There are many other factors that
go into prioritization of a signal. On this screen, I, or slide, I have given you 3 examples
of, well 2 practical examples, one from the WHO, who use an ER based triage process in
their prioritization process/ And, again, for this same reason: to ensure
that they use appropriate resources for appropriate signals or potential signals we should say
at this point. And then the NHARA uses a different analytical method and it’s used as, it, it
is comprised of mathematical scores that will calculate themselves and contribute to an
overall score for the signal and that score is what drives whether or not to use that
signal into the next phase of signal management, which is in fact, signal evaluation/ There are, however, other articles in the
literature and I have given you a few references in the last part of this presentation that
have very interesting and very slick ways of prioritizing various signals that may occur.
Notably, an article from Johnson and Johnson, where they have a very comprehensive method
of formally prioritizing and then evaluating each of the signals that they detect. I went
through manual investigations, through case safety reports, or periodic PSUR reporting
or through automated signal detection such as Empirica signal. Finally, signal evaluation. This is the formal
process of reviewing scientific data sources to either refute or confirm the existence
of the signal within your company product safety profile. And this is what will feed
the signal into the risk management process at your company. So, this should be a very multi-faceted approach,
according to CIOMS 8 working group. You may choose to collect evidence to evaluate this
causal link that you’ve identified or the potential causal link. You may utilize these
public databases to gather background rates to see if the potential signal that YOU are
looking at warrants further investigation. There’s many ways to accomplish this signal
evaluation component of signal management. And while I allude to it, I thought I would
show you the actual CIOMS 8 working group signal detailed signal management lifecycle.
I tried to simplify it by categorizing in the arrow, the large arrow boxes, those activities
of the CIOMS 8 working group that are related to one another in my modified simple form
of signal management, which really just comprises itself of three safety properties: signal
detection followed by sign prioritization followed by signal evaluation. So, as you can see, there is actually a lot
of detail within each of those simplified steps within signal management. And all of
that is clearly outlined by the CIOMS 8 working group in a very easy to understand manner. For today’s discussion, we’re really focusing
here on the signal evaluation step within the signal management lifecycle. It is within
this step that we may choose to use pharmacoepidemilogic study designs to, again, confirm or refute
various signals that may have been identified. In this small table that I am showing you
right now is in order of difficulty of execution and increasing your strength of causal evidence
from the bottom to the top are the various types of study designs. Some are observational
and some are experimental. As you can see, we go all the way down in case reports, which
would be the far easiest of the study designs to execute but it has the least amount of
causal inference capability. And as you transverse upward in this table, you will gather your
strength in causal evidence, but you will increase the order of difficulty in executing
these types of study designs. The experimental study design that we have
here is at the top, and that would be your randomized, typically, your randomized control
trials. Potentially you could also have experimental designs that are not controlled. Just a small side note, as you can see in
the previous table, there has been a discussion of case control studies as well as cohort
studies. And, there is sometimes some misunderstanding on the differences between these two types
of epidemiologic study designs. So, this slide, I thought, would be very helpful in, at least
it helped me let’s say, in understanding the differences between the case control and the
cohort study design. And really, the root of this discussion is what do we know what
do we want to know. So, for example, if I know a group of individuals have a particular
disease of interest that I am curious to know about the exposure of our medicinal product
and its relationship to that disease, then I probably need a case control study design. If a different research question in the prospective
or retrospective cohort study design, where we know certain groups of people’s exposure
to our product, but we’re unaware of the disease of interest. Which have it, which do not.
So this is really the key difference between a case control and a prospective or retrospective
cohort design. There are many reasons for pharmacoepidemiologic
studies. There are regulatory reason for doing this, many now are required for approval,
especially in identifying risk management processes. You may have several pharmacoepidemiologic
studies as a response to a particular audit. They can also be a very powerful tool in marketing.
They can assist in market penetration by further documenting, a safety profile. For example,
comparing your product to other products on the market, either like minded or different
chemical entities. It can also be used as a name recognition increases. For example,
you conduct these research efforts, you publish the articles, you increase the brand recognition
of your product. There are many examples of repositioning the
drug based on these pharmacoepidemiologic studies in the literature. This can have a
wide effect. You may find age, gender, ethnic or genetic differences in a particular patient
populations. This can then aid in the targeting of the drug to particular participants. You may have different, umm, outcome measures,
so, for example, quality of life aspects that you like to see in the marketplace. Does our
product increase quality of life? And there is always the wonderful opportunity, hopefully,
of the medicinal product or the product in nature, creating an unintended benefit of
the product, which then could lead to future indications or multiple indications of use. Unfortunately, in our world, we also are faced
with legal reasons that we need to do these pharmacepidemiologic studies. This could be
either in anticipation of or in response to particular pending legal action. You also may have clinical reasons that you
want to perform these pharmacoepidemiologic studies. You may be trying to generate new
hypotheses, to increase particular knowledge of a particular part of your safety profile.
This is especially useful when your product is a new entity in the marketplace. And you
may actually want to do some hypothesis testing, looking at particular benefits of the product
and its effect as well as may be even harmful ones. Again, this is where we would look toward
the pharmacoepidemologic observational study designs to actually test particular hypotheses. And all of these examples are very clearly
noted in a comprehensive text book by Strom & Kemmel as it is also an excellent text to
have. I have given you the reference in this slide deck as well. That is a much, much larger
textbook that is a formal academic textbook. And it’s a little harder to get through, but
it is far more comprehensive than Waller. Alright, so I do again apologize, but we are
limited with time, so that was a very quick overview of Pharmacoepidemiology and some
of the study designs that we utilize. And now I want use all of that to give context
to the next two sections of the slide deck. The first, we are going to look at 2 very
exciting free online databases, where you can perform a number of these type of activities
or confirming or refuting your signals that you may have detected and formally moved into
evaluation. So, again, there are two broad kinds of databases
that we use in this type of work. There are for FREE databases and then there are for
FEE databases. And I would say that them the for FEE databases are far more known, utilized
and published on than for FREE databases. But the for free databases, at least the two
that I will demonstrate to you, could potentially give you the access to some information where
you did not have to pay large fees in order to gain the data in order to accomplish these
pharmacoepidemiologic trials. So, the two that I am going to focus on that
are the for FREE databases are the CDC-WONDER database and a brand new database called the
EU-ADR. So, the CDC-WONDER is obviously very US focused. All of the information comes from
patients found in the United States. EU-ADR is very EDU focused and the patient’s data
that is found in that database is obviously European Union Based. So all of the biases
associated with the data collection between US based patient populations and EU based
patient populations holds for those two databases. So, please make sure you’re aware of that
before you start to draw inferences from the information that you carry. So, the CDC-WONDER database is a wide ranging
online database for epidemiologic research. It is a very easy to use internet based web
tool provided to us by the CDC, the Centers for Disease Control and Prevention. It is
actually available not only to public health professionals at large, be even the public.
And it has a huge amount of information from multiple data sets governed by and reported
from the CDC. Once you attach yourself to that website,
you will see a number of public databases that you’ll be able to query on. One of the
major limitations with the CDC-WONDER is that we are limited to only a mandatory reported
event. So, this would be death events, cancers, HIV/AIDS, STIs, a lot of information on sexually
transmitted infections or STIs. So, if your medicinal product has any of those for therapeutic
uses, the CDC-WONDER databases will provide you a very nice interface to gather the information
about prevalence and incidence rates in the US population for your various diseases of
interest. If, however, your product of interest is outside
of those areas of indication, you may not find the CDC — WONDER website useful. I’m
going to, for the next few slides, assume that not 100% of the sub-population afflicted
with the disease dies from that disease. And this means that we may able to use the mortality
rate as a confirmation or refutation of a suspected ADR. So, for example, we’re going
to use the CDC-WONDER’s mortality database to look at the causes of death that may have
some type of relationship to a disease of interest that we have found in our safety
signal, either using our safety individual case safety reporting system such as Argus
safety or using a signal detection system like Empirica signal. So, to do this, I access the mortality database
and when I come in here, I am able to associate and look at the information in various ways.
So here, for example, and can group my results by age group, if that is appropriate to me,
I can print out optional measures, I can look at crude rate/standard error, I can look at
a percentage, a raw percentage of total, or even a crude rate/confidence interval. One of the differences you’ll see with the
CDC-WONDER database is much like of the Department of Health and Human Services, the information
is coded to the ICD dictionary and not to Medra. I can also focus my query on the mortality
data to particular states if I wanted to or I can choose the entire United States. I can
then, again, take particular age groups, genders, race, ethnicity as defined by the US government
for the organization. How close are they to a metropolitan area or not. I can then look at particular year and month
that are captured in the mortality database. And all of this, what you choose, what shouldn’t
you choose, is all decided up front in standard research question development, right. So,
I know what to select here in my query because I have spent time with my Biostatistician
and my medical directors working out the hypotheses of the type of query I need to get the answers
to. You also have the autopsy information, if
that is appropriate, again for your research questions. So, many, many options here in,
just in the mortality database alone. Once you’ve added all of the parameters around
your query, the next step would be to select one or more of your causes of death. And,
again, these have been coded, all the deaths that have been reported to the CDC, have been
reported back and coded into ICD-10 codes specifically. So, now I would focus on the particular malignancies
that is appropriate for my research hypothesis or question that I’m asking, which in this
case, I’m looking at what is the underlying incidence rate in the American population
for a particular type of malignancy, in this case, I’m looking at neoplasms of the stomach. I can then select my calculated rate. So,
how do I want this data to display and per 100,000 is the typical way that mortality
rates are would be displayed so that is what I have selected here. And you have various
other options in the other options section. Once I have all my attributes done, then I
can click the send button and I will receive essentially a table listing out my crude mortality
rate per 100,000 individuals in the resulting table. And now I have a comparative measure
to my rate that I’m seeing in my signal that I have detected either using advanced signal
detection algorithms like Empirica signal or in my traditional method of individual
case safety or periodic reporting analysis. Okay, again, we did make a number of assumptions
to get to this point, I find, at least for this example, I believe that the assumptions
are valid. You would have to, again, work with your Biostatistician, your medical directors
and if you lucky enough to be a part of a firm that has a department of epidemiology,
your epidemiologist on staff to make sure that all of the assumptions, hypotheses and
information that you are gathering is appropriate to the research questions you’re being asked. Again, this is just another example of one
database within the CDC-WONDER system. There are many, many other databases within the
CDC-WONDER system that may be advantageous, may pose an advantage to your organization.
So, even if the mortality rate is not a good example for your company, I strongly advise
you to go out to the CDC-WONDER website and look to see if there’s any other database
that may be helpful in estimating prevalence or incidence rate within a particular population
for your company. Okay, the next to a FREE database that I would
like to look at is this really amazing online database that I stumbled across last year.
And it’s called the EU-ADR project. It’s extremely innovative. It is using computerized systems
to detect adverse drug reactions that will in turn supplement our spontaneous reporting
system such as Argus safety or even Empirica safety. I’m sorry, Empirica Signal. So, once,
this is a relatively new database project, so certain pieces are still being worked on.
But once it is complete, it will have electronic health records not just electronic patient
records from individual case safety reports, but actual electronic health records from
over 30 million patients from several EU countries. This uses a very interesting and complex method
of text mining and other epidemiologic and computational techniques to analyze all of
those various health records to detect signals and then rank those signals back to you in
a very easy to read manner. This does require you to register for access to this website,
but it does not cost. Once you are registered, you can log into EU-ADR website, (coughing),
excuse me, and you’ll be, you’ll be possibly using two main tabs: A data set tab and a
work flow tab. The data set tab is where we go to create
particular drug and event pairs that we’re interested in. And these contain all marketed
drugs listed the WHO big drug dictionary. And all of those drugs are coded all the way
down to the ATC level 5. This is a, an uncommon level to code our WHO drug verbatim terms
all the way down to level 5, but level 5 actually is the medicinal product record for a particular
drug within the WHO drug dictionary. So, on this screenshot you can see these J-numbers
in front of everything, J04AM02. That represents in the WHO drug dictionary a very specific
drug. Maybe a combination medication, it may be an individual medication, but it represents
a single chemical entity construct. Then you have the ability to select a particular
disease of interest or event. And these events here are coded in this particular screen to
specific abbreviations that are specific to the EU-ADR web-based system. So, again, not
ICD-9, not ICD-10, not Medra, it’s a proprietary event code specific to this particular systems.
But once you created this drug and event combination, you can now path the drug event pair into
the various engines of the comprehensive searching database. The first database that you can put the event
drug pair into being a review of all of the Medline literature that is available. And
here on the screen I have given you the algorithm that is utilized when it is querying all of
the publication data to decide whether or not to return to you a signal score. So, certain
things have to be met in the literature in order for this particular database to return
you an indication that this event-drug pair is in fact a signal. There is also a Medline
co-occurrence engine that you can run this event and product pair into. You can look
at a daily media search engine as well, which is a separate publication engine. All of these
work in the same way: they take your drug and event pair, they search the available
literature and if certain things are met in the logic of the algorithm then it returns
to you a potential indication of a potential signal. The Drugbank database is a separate database
maintained by a separate university in the Netherlands. And, again, works in the exact
same way. It passes the drug and an event pair into that database, looks up all of the
available data with respects to that drug-event pair and if particular thresholds are met,
it will return an indication of a signal. The substantiation engine is fascinating in
my perspective. It takes genetic information, again maintained by a particular organization
in Spain, and it passes that drug-event pair into that engine and it looks for a genotype
that are actually related to exposed phenotypes of your event to see if, in fact, there is
a causal biologic or genetic relationship between your even and your product pair. This
is a really exciting and very, very interesting area for signal detection and can really aid
in not just confirming a signal but can also introduce to you particular pieces of literature
that may be used in your confirmation or refuting of a particular signal. Again, when path each of these drug and event
pairs into the separate databases we just described within the EU-ADR web based system,
you will either find results or you will have these warning messages that tell you that
there were, in fact, no retrieved relationship. There are some definite limitations to this
database. First of all, it is only from a select number of EU countries, so generalized
ability is not a possibility with that. The system doesn’t really provide a fantastic
way to of coding your event turns, this is a very specific way of doing that. So, you
really have to become familiar with the three letter even representation. But it does have
this incredible ability to substantiate based on biologic plausibility within human genomics. Okay, so again, apologies for the brevity
I hope again this is just an overview of these databases, once you get the slides and the
references, I really strongly advise you to go out there, play around with it, if you
have any questions, please feel free to email me about them. I will try and answer questions
as you have them. So, now let’s look at the traditional, the
more traditional for FEE based databases, which are arguably far more substantial in
their data. They are arguably better organized, better equipped to handle these pharmacoepidemiologic
type trials. So I’m just going to cover three. And, again, I’m going to cover them very quickly.
These are three of the largest that you may have already seen or heard of. The Group Health
Cooperative, Kaiser Permamente Medical Care Program, and the UK Clinical Practice Research
Datalink, which you may all be more familiar with as the GPRD database. It has now been
rebranded. So, let’s first look at Group Health. The
Group Health Cooperative is a non-profit healthcare system that is essentially a large HMO. And,
as a large HMO, They have access to electronic medical and health care records for particular
patients, subjects. In fact, it’s about 600,000 people in Washington and Idaho and has a number
of automated and manual databases and those datasets can serve many different of apidemiologic
studies. And this is a fairly stable set of patients and populations over time. So, that
goes to adding to the credibility of the information that you gather from such a dataset. I did, in each of the summaries of these four
fee-based, I did a description of the database, some of its limitations, and then an example
of the use that database. And these are all taken directly from the Strom and Kemmel with
the exception of the GORD, which I have rebranded to the CRPD, which is its current nomenclature. So, here, for example, I give an example of
the retrospective cohort studies that were published from the GCH Cooperative databases
that looked at perinatal outcomes of congenital malformation and early growth and development
of infants with and infants without prenatal exposure to antidepressants. So, this is very
much what we were talking about before. A retrospective cohort study where, once analyzed,
we saw or the scientists of this article saw, that discharge, by looking at the discharge
data, the pharmacy data all of the exposure that was present in the large comprehensive
GCH databases. They were able to match infants exposed and not exposed to the particular
antidepressant of interest and they blinded the medical reviewers themselves. So, the results of this indicated that tricyclic
exposures did not have an association to any of the outcomes of interest which were premature
birth and lower delivery weights. So, these types of databases, again, can be used to
perform these type of cohort studies right within the datasets themselves without recruiting
patients or subjects and following those subjects up over time such as Sidenow Safety Servant
Surveillance studies that maybe we’re more familiar with, which as some of you can attest,
are quiet expensive to do and take time. So, access to these for fee databases can really
substantiate your signal detection methods and may, even though they cost money as well,
may provide a more cost effective manner than again creating large prospective cohort studies
designs like Sitenal site. There are some limitations to the GCH database
itself. The size of the database really wouldn’t allow you to indicate rare or identify rare
drug AE combinations just because it’s a smaller, even though it’s a larger repository, it’s
still, in terms of size, still a little bit smaller than some of the others. So, let’s look now at the Kaiser Permanente
Medical Care Program. This is definitely one of the largest US based non-profit HMOs. It
has over 8.2 million individuals and covers 8 states. This is an enormous for fee database.
And this can definitely empower researchers to do some really large scale observational
trials. One such example was a retrospective cohort
study of patients exposed to Rezulin in an attempt to understand, identified hepatic
failure, issues that came up with that drug back in the ’90’s, This cohort was 9,600 diabetic
patients who were exposed to the Rezulin product over three years. The researchers looked at
the hospital discharge summaries and procedure documentation that indicated any type of hepatic
injury and they identified approximately 1,000 individuals records after the review. And
109 of these were sent to a blinded panel of heptologists and outcome, for outcome adjudication
rather. The blinded panel actually only identified
35 cases where hepatic was attributed to the use of the diabetic medication. However, the
entire diabetic population did have an increase of hepatic injury compared to a general population
of patients. There, again, this is information we would
not have been able to identify without access to these large datasets where we could, for
example, compare outcome rates of hepatic injury in one group compared to another. So, even though you looked at the FDA AIERS
database using the Empirica Signal, if you queried the Rezulin product in that particular
database, you would see that it has a disproportionality score that indicates a 20 to 25 times higher
experience of hepatic failure in Rezulin subjects versus any other product and subject in that
database. This particular observational study disputes that finding and really suggests
the hepatic failure in the Rezulin user is far, far smaller. However, we should put that
into context. We can look at the CDC-WONDER database in that very same mortality rate
for hepatic failures not elsewhere classified and we see a very, even a much smaller scale
of particular hepatic issues within the CEC-WONDER mortality database. So, again, these observational trials can
be very powerful in helping refute or confirm identified signals that may come out of these
large signal detection databases such as Empirica Signal. Again, due to time, I am not going to go over
some of the limitations, but there are a number and all of those are very important when you
are analyzing and reporting back on the analyses. So, you do need to make sure that you do understand
those limitations of the data mart that you’re using. Alright, so now let’s just focus on the UK
Clinical Research Practice Datalink, also known as the General Practice Research Database
or the GPRD. This has datasets comprised of both administrative records and patient care
records. The administrative records are often used in the billing cycles and they may not
have accurate diagnosis data. It depends on the research question that you’re asking as
to whether or not you should include the administrative records in your analysis or not. But, they
do have some information in there that could be useful depending on your research question
that you’re formulating. And, then secondly, there is patient care
data and these are all the individual medical records used in the allopathic care of the
individual patient. All of their collection and storage, again, may not be appropriate
for your research question in an observational study design, but you really need to work
with your epidemiologist and biostatistician personnel to flush all of that out ahead of
time. In the UK, the CPRD is really considered to
be one of the world’s largest databases in use. These were many years ago called the
Value Added Medical Products Research databank in 1987 and has been adding approximately
3 million patients per year into the database ever since. And all of, 1 million of these
patients have more than one, more than 11 years’ worth of observational data over time.
This is a huge advantage in giving us the ability to perform some very interesting observational
studies models. One example is a retrospective cohort study
that looked at acne patients from 1987-2002 who had been exposed to antibiotics and those
who had not. The outcome measures that were looked at were any upper respiratory infections
of a 12 month period. And all of the results were adjusted for age, sex, year of diagnosis,
number of prescriptions for acne, office visits, history of diabetes and asthma. All of these
things were considered to be potential confounders for the outcome of measure, which again, were
upper respiratory infections. The results of this study indicated that the
acne patients exposed to chronic antibiotic treatment did, in fact, have and demonstrate
an increased risk of upper respiratory infection, almost 2.15 times that of anyone not exposed
to chronic antibiotic treatment. And this did not change once the confounders were added
in. Unfortunately, the etiology or the origin
of the upper respiratory infection was not evaluated, i.e. whether it was bacterial or
viral and we also don’t know if acne patients in general are more prone to upper respiratory
infections independent of antibiotic use. But, nonetheless, these are, is a very interesting
study design and something capable within these large observational data sets. And lastly, again, there are limitations within
the CPRD database itself. And, again, like all of the other databases, you need to be
very aware and cognizant of those limitations in your interpretations of the results and
any type of extrapolation to your population or to a general population, a more generalized
population than you are working with. Again, very sorry about the brevity, but I
wanted to leave time for questions. Here are the references. You will get all of these
in the PDF version and again, you have the recorded webinar to refer to as well. I think now what we can do is open the webinar
up for particular questions. Great. Thank you Rodney. So, before we move
onto the Q & A sessions, I want to remind everyone that you can ask questions via the
chat feature. Try to be a clear as possible. And we’ll get to as many questions as time
allows. Rodney, I think it may make sense for you to take a look at the questions tab.
There are several questions that came in and due to as many acronyms as there are it might
make sense for you to just go ahead and deal with those exact questions in case I (both
taking) completely. (Rodney): Yeah, no problem. So, I’m just going
to start from the top and read the questions out. So, the first is: Where do the population
numbers come from? So, I’m, I’m not sure which of these databases this question is in regards
to. But, in general, that information is usually found on the limitations slide in my slide
deck. Bu, for example, the CDC-WONDER data are from mortality events, in the mortality
database that are deemed mandatory reportable events to the CDC. Those come from all death
certificates, etc., in various county and state corner offices. The STI data that comes
in comes, again, as reportable sexually transmitted infections like syphilis, gonorrhea, Chlamydia,
et cetera, HIV. Those are all mandatory reports from an individual physician or to a county
health department official. So, really just depends on the database as to where the data
is coming from. And there was a clarification of what that
question was in regards to CDC-WONDER. So…… Okay. Great. So, I hope that answered it. Okay, the next question is: Where did the
EU-ADR includes medical device AE pairing? Yes, this is a very interesting question.
I have not reached out to them yet, but I plan to, to ask them make sure that they do
include that. It is not on their horizon right now because, again, their focus right now
is on the WHO dictionary and I assume the person that asked this already knows medical
devices are not encompassed in the WHO drug dictionary. So, it’s a limitation right now
in the software. But, hopefully in the future we will have something similar. But once I
reach out to them I will keep your name and send you an email and let you know what I
find. Okay, the next question is: If the events
do not come from a standard library, how are they determined or how do you, how do you
find them? This is a great question and I assume also that this is in regards to EU-ADR
system. That is explicitly stated in their user guide. So, you will, you will have access
to the small table that shows you the various 3 letter codes that represent certain disease
states. I can only assume that over time as more drug pair and event data get added to
that system, that the 3 letter codes will increase over time as well. So, I shouldn’t
have made it seem like it wasn’t going to be comprehensive. It is very comprehensive.
It’s just, it is a different type of event coding that what we are used to in the safety
world. We typically would code to ICD-9 or 10 or MedDRA. The next question: We understand your focus
today is on evaluation rather than detection. But we are interested in what database and
tools you might recommend for signal detection, i/e/ for adverse reactions in marketed products?
Sure. So, again, I think the EU-ADR system is a wonderful tool to use for free that will
give you a, the capability of identifying potential signals within a particular event
to drug pair, although as we have already discussed, it’s quite limited in its data
right now. But, that is an ongoing project and it will expand over time. For now, the
products that are available on the marketplace for these large scales queries of databases,
public or private databases, are really one the Empirica signal tool from Oracle. And
that tool queries the WHO disease base and FDA AIERS datasets and displays to your potential
scores for disproportionate ratio, which would indicate to you a potential signal there.
There’s also now a new product by oracle called the Empirica Healthcare which looks at similar
information, but from these large robust healthcare databases like GPRD and others. Much, again,
for the purpose of signal detection. The next would be, the question is: Please
suggest a vaccine ADRs database other than VAERS. Yeah, that’s a good question. Unfortunately,
the only vaccine database that I’m currently aware of and using actively is the VAERS database.
I will have to get back to you on alternatives. But I do have your name here and I can follow
that back up with you. The next question is: What is the price to
get access to GCH in order to evaluate, for example, couple of variables or outcomes?
Yes, so, I don’t have any pricing information for these for fee databases. You must contact
the vendor specifically which in this case is GCH system. So, all of the contact information
for these for fee are in the reference section. So, you should be able to reach out directly
to them. The next question is: Are the datasets that
you could access via the CDC-WONDER considered clean or raw? That’s an excellent question.
They are not the raw. They are clean. They have been entered into various databases and
they have been coded. So, that doesn’t mean that the raw equivalent is not available in
the data sets. So, for example, description of the event as recorded versus coded data.
They could be available there, you, we would just have to look to see. But, you should
probably consider them clean. And then, I have a final question here: Have
you encountered the situation where a new indication of a marketed medicinal product
was granted based on the outcome of an epidemiological study only? That is what I understood from
the presentation but it may not be true or very likely. Thank you. Yes, sorry if I gave
that as a, as a, um,….I’m sorry if I made it seem like that I had seen that. No, I have
never seen that. Drug approval has in my experience always come from randomized controlled design.
However, again, observational study designs aren’t without merit. They can give credence
to new indications of use. But, that’s more hypothesis generating than hypothesis testing
that would feed into a, let’s say, a MBA submission. But, I should preface that with, that my experience
in pharma is, I have a number of years in it, I’m not used, I’ve done two submissions
in my entire career. So, I’m hardly the expert with respects to all possible submissions
that are out there in the world. So, please take that with a grain of salt. But, yes,
from my perspective, I’ve only seen NVAs granted from randomized clinical trials. Somehow we have ran out of time, so thank
you all. We don’t have any more questions at the moment.
I do want to remind everybody that the presentation and recording will be sent today or tomorrow.
So, please look out for that. And, if for some reason you did not receive my email,
you can always check on our website for the recording as well as other upcoming and past
webinars. If, you have any other questions after the
webinar, feel free to contact us via phone or email. Rodney’s information is on the slide
right now. You can also email us at [email protected] We want to thank you very much for your participation
and hope that the information that Rodney provided was helpful. And we hope that you
have great rest of the day and evening. Thank you so much for joining. Thank you everyone. Have a good day.

Leave a Reply

Your email address will not be published. Required fields are marked *