Russ Altman: What really happens when you mix medications?
Russ Altman - Big data techno-optimist and internist
Russ Altman uses machine learning to better understand adverse effects of medication. Full bio
and get some tests.
that you have high cholesterol
from medication to treat it.
that this is going to work.
a lot of studies, submitted it to the FDA.
skeptically, they approved it.
of what the side effects are.
of a conversation with your physician
because you've been blue,
in life quite as much as you usually do.
I think you have some depression.
you another pill."
about two medications.
of people have taken it,
the FDA looked at it -- all good.
these two together?
if bad things are happening
who has several diagnoses
and really, in my opinion,
to understand these interactions
of different sources of data
when drugs can be used together safely
because that's his name.
to understand how drugs work
and how they work separately,
an amazing database.
download it right now --
of adverse event reports
that the patient has,
or side effects, that they experience.
that are occurring in America today,
of thousands of drugs.
and we know it's involved with diabetes.
look at the side effects of a drug
is likely to change glucose or not."
that were known to change glucose
that don't change glucose,
in their side effects?
In urination habits?"
to give him a really good predictor.
with 93 percent accuracy
you have to build his confidence.
knows all the drugs that change glucose,
but not really that interesting,
I thought you might say that."
so I did one other experiment.
who were on two drugs,
did not change glucose,
Good idea. Show me the list."
not very exciting.
was, on the list there were two drugs:
a cholesterol medication.
of Americans on those two drugs."
at the time, 15 million on pravastatin,
with their glucose
that he did in the FDA database
with the mumbo jumbo,
evidence that we have."
electronic medical record.
that's OK for research,
on these two drugs
and thousands of people
that take paroxetine and pravastatin.
and had a glucose measurement,
another glucose measurement,
something like two months.
we found 10 patients.
had a bump in their glucose
we call this P and P --
the second one comes up,
20 milligrams per deciliter.
if you're not diabetic,
about a potential diagnosis of diabetes.
don't have a paper,
and -- give me a break --
at Harvard and Vanderbilt,
Vanderbilt in Nashville,
medical records similar to ours.
the glucose measurements
in one week found 40 such patients,
from three diverse medical centers
getting these two drugs
we had left out diabetics,
have messed up glucose.
at the glucose of diabetics,
per deciliter, not just 20.
"We've got to publish this."
was in review, went to the lab.
who knew about lab stuff.
but I don't do pipettes.
one P, paroxetine.
of mice both of them.
20 to 60 milligrams per deciliter
based on the informatics evidence alone,
if you give these to mice, it goes up.
could have ended there.
thinking about all of this,
of it, but somebody said,
who are taking these two drugs
one new medication or two,
or the one drug you're taking,
their search logs with us,
these kinds of searches.
denied our request.
who works at Microsoft Research
the Bing searches."
companies in the world,
to make him feel better.
you might not understand.
to do searches at Google,
for research purposes only."
my friend at Microsoft.
that a regular person might type in
"urinating a lot," "peeing a lot" --
of the things you might type in.
that we called the "diabetes words."
that about .5 to one percent
involve one of those words.
or "Paxil" -- those are synonyms --
of diabetes-type words,
that there's that "paroxetine" word.
to about three percent from the baseline.
are present in the query,
that we were interested in,
or hyperglycemia-type words.
their side effects indirectly
to the attention of the FDA.
for doing this, and others,
either individually or together,
Why tell this story?
of big data and medium-sized data
a new ecosystem
and to optimize their use.
at Columbia now.
for hundreds of pairs of drugs.
very important interactions,
is a way that really works
of drugs at a time.
on three, five, seven, nine drugs.
to their nine-way interaction?
A and B, A and C, A and D,
D, E, F, G all together,
more effective or less effective
that are unexpected?
for us to use data
the interaction of drugs.
that we were able to generate
volunteered their adverse reactions
through themselves, through their doctors,
at Stanford, Harvard, Vanderbilt,
and security -- they should be.
that closes that data off,
and it was a little bit of a sad story.
the two drugs very carefully together,
when you're prescribing.
two drugs or three drugs
of causing a side effect,
for depression, for diabetes --
TED Talk on a different day,
of drugs in combination
of our patients even better?
About the speaker:Russ Altman - Big data techno-optimist and internist
Russ Altman uses machine learning to better understand adverse effects of medication.
Why you should listen
Professor of bioengineering, genetics, medicine and computer science at Stanford University, Russ Altman's primary research interests are in the application of computing and informatics technologies to problems relevant to medicine. He is particularly interested in methods for understanding drug actions at molecular, cellular, organism and population levels, including how genetic variation impacts drug response.
Altman received the U.S. Presidential Early Career Award for Scientists and Engineers, a National Science Foundation CAREER Award and Stanford Medical School's graduate teaching award. He has chaired the Science Board advising the FDA Commissioner and currently serves on the NIH Director’s Advisory Committee. He is a clinically active internist, the founder of the PharmGKB knowledge base, and advisor to pharmacogenomics companies.
Russ Altman | Speaker | TED.com