state_sleep_2022_1012

December 12, 2022

ID9273

To CiteDCA Citation Guide

00:00Sleep centers and just a few announcements
00:05before we introduce today's speaker.
00:08So first, please take a moment
00:10that you are muted and if you
00:12wanted to receive CME credit,
00:14please see the chat room for instructions.
00:17If you're not registered yet,
00:18you'll have to do that
00:20before you receive credit.
00:22And then usually the recording of
00:24this session is available online
00:26within a couple of weeks and the
00:28link will be provided in the chat.
00:30And if you have questions,
00:32please make use of the chat room.
00:35And throughout the hour.
00:37And we will get to them at
00:39the end of the talk.
00:40And so I'm going to hand it
00:43over to Doctor Javaheri,
00:45who will introduce today's speaker.
00:48Welcome, Sir.
00:49Well,
00:50thank you so much.
00:51It is my honor to introduce
00:54Doctor Ali Azar Barzin.
00:57He's an assistant professor in the
00:58Division of Sleep and circadian rhythm
01:00disorders at Harvard Medical School,
01:02and he's director of the Sleep apnea.
01:05Health Outcomes Research Group
01:06at Brigham Women's Hospital.
01:08He's written multiple algorithms to help
01:11improve characterization and diagnosis
01:13of obstructive sleep apnea and to
01:15help identify patients at heightened
01:17risk of morbidity from sleep apnea.
01:20He's really paved the way for precision
01:23medicine and sleep apnea treatment.
01:25He has many publications and and grants
01:29including two R1 and R2154 original peer
01:32reviewed papers and high impact journals.
01:34So.
01:35It's really a treat to have
01:37him join us today.
01:39And without further ado,
01:40I'll hand it over to you, Ali.
01:43OK. Thank you so good for inviting me
01:46to share my research and thanks for
01:49thanks everyone for attending my talk so.
01:54Can you see my slides and hear my voice?
01:58Yes, but you, yeah, and it's in presentation.
02:02OK, great. So the title of my talk
02:06is understanding sleep apnea beyond
02:09the apnea hypopnea index or HIV.
02:13So this slide shows my disclosure.
02:16I just give you a few seconds to.
02:19Taking note of the numbers if you want.
02:22And then I'll continue.
02:28OK, great. So I'll add the the this
02:31slide is actually the outline of my talk.
02:34I'll talk about the background.
02:37And the rationale for?
02:39Developing new metrics,
02:41you know, in sleep apnea.
02:43And then I'll review the hypoxic burden
02:46and and and talk about its determinants.
02:50Then I'll talk to you.
02:51I'll talk about the heart rate
02:53response to respiratory events and
02:55it's importance and if I have time.
02:58At the end of my talk,
02:59I'll talk about the EEG arousals and
03:02their relevance for risk prediction.
03:07So this, I'm sure many of you know, you know,
03:11obstructive sleep apnea better than I do,
03:13but I just quickly go over.
03:16The different signals that
03:18we collect during PSG.
03:20So this shows a 5 minutes PSG tracing.
03:25Of obstructive events for a
03:28patient that have service apnea.
03:30And as you see with the absence of air flow.
03:36There is an out of phase
03:39movement of thorax and abdomen.
03:41And the reduction in oxygen saturation?
03:45The end of each event is accompanied
03:48by an arousal from from sleep which.
03:53Then recovers the SP O2 back to normal.
03:58Levels and.
04:00The transitioning back to sleep
04:03is associated with collapse
04:05of upper airway and recurrent
04:08obstructive apneas during the night.
04:13OK. And and we use a metric called
04:15apnea hypopnea index which is
04:17the number of apneas or complete
04:19collapse of the airway and hypopneas.
04:22Of the events with partial obstruction
04:24in the airway per hour of sleep.
04:28As you know, sleep apnea is associated with
04:32many comorbidities and and health outcomes,
04:35including excessive daytime sleepiness,
04:37reduce quality of life,
04:39impair workflow performance,
04:42increase car accident risk and
04:46several cardiometabolic outcomes.
04:48And for example, in hypertension,
04:51if you define OC as an HIV
04:53AIDS in 5 minutes per hour,
04:55the prevalence of OSA in
04:57individuals with hypertension.
04:58About 82% and it's about 30% if you
05:02use HIV AIDS and 15 events per hour.
05:05And similar associations with other.
05:09Cardiovascular diseases.
05:14However. Despite that,
05:16observational studies of CPAP
05:18showed that CPAP is beneficial.
05:21There have been several recent randomized
05:24control trial that have not been able.
05:27To detect a benefit of
05:29CPAP including save cuts.
05:31And an Isaac study from Spain.
05:35So in addition to steep adherence
05:39and other problems with the
05:42randomized control trial of of CPAP.
05:46An important reason for for now
05:48finding in in these R cities,
05:51we believe is inadequate patient selection
05:53by by indices like apnea and Putnam index.
05:57And the reason for that is HIV is a
06:01simplified metric is a frequency counter.
06:04And doesn't capture the magnitude
06:07of physiological changes that
06:09occur during sleep.
06:10And and for any given HIV,
06:13there are radiation in in
06:16event characteristics.
06:17There are differences in the air
06:19flow reduction in the number
06:21of apneas versus hypopneas.
06:23There's desaturations severity.
06:25We're missing the depths and duration
06:28of Desaturation and we also are
06:31not measuring the cardiovascular
06:32and EG acute cardiovascular and EEG
06:35responses to respiratory events
06:37including changes in heart rate,
06:39blood pressure or or egg characteristics.
06:43How about other commonly used
06:45metrics of our system variety?
06:47They unfortunately they
06:49share similar limitations.
06:50For example oxygen desaturation index.
06:53Again it's.
06:55It's a frequency counter,
06:57doesn't have the depth or duration
06:59information. Minimum saturation.
07:03It's just oversimplifies the dynamic
07:06of breathing and the saturation to just
07:09one number number for 8 hours of recording.
07:12Arousal index again is is a
07:16frequency counter doesn't consider
07:19the intensity of arousals.
07:21And one of the the better metrics is T-90,
07:25which is basically percent time below 90%.
07:29However it's based on arbitrary
07:32threshold of 90%.
07:34And it's not specific to OSA completely,
07:39it depends on baseline SPO 2 and.
07:44And and if there are other
07:48comorbidities that will affect the
07:50the value of T-90 so it may not
07:53respond to CPAP treatment.
07:56So it's the the problem with T-90 is
07:58that it's not specific to sleep apnea.
08:03OK, so due to all these limitations
08:07of commonly used metrics,
08:10we need to better quantify
08:12event characteristics.
08:14What do you see on the right side?
08:17Is the uh. So we we have time zero.
08:21Is the end of events and we
08:25just overlay all the events.
08:27These these Gray lines as you see here
08:31and and we took an ensemble average
08:34which is the red line in the top.
08:37Top row you see ventilation leader per
08:40minute and then you have the SP O2 signal.
08:43In the 2nd row you have heart
08:46rates and systolic blood pressure.
08:48In this this panel I'll I'll
08:50talk about how this ensemble
08:52averaging works in the next slide,
08:55but as you see here,
08:56there are.
08:57A large variation in the events
09:00characteristics and both within and
09:03between subjects and with with apnea
09:05hypopnea index and other metrics we were
09:09missing all this important information.
09:13Information like events related events into
09:16deficit which we call it when something
09:19better than event related hypoxemia,
09:22hypoxic burden.
09:24And eventually the changes in heart rate,
09:26blood pressure which can be used
09:29to quantify autonomic arousals and
09:32event related changes in each G like
09:35metrics like arousal intensity.
09:38So we are missing all this
09:40information from PSG.
09:42So I'll be talking about hypoxic
09:45burden in this next section.
09:47So this this graph we have air flow.
09:53Here on the top and then
09:56we have tidal volume.
09:58We have SP,
09:59O2 signals and in EKG at the bottom here,
10:03so I'm just going through the algorithm
10:06and how we developed algorithm.
10:09And the marking on the top the the red.
10:14Rectangles. They're hypopneas.
10:18That were manually scored
10:21for this period of the PSG.
10:25So one way to measure the burden
10:27of OC is to just measure the area
10:29under the desaturation curve.
10:31As you see here.
10:35But sometimes it's difficult to really
10:36find the start and end of this situation.
10:39So we because of that we
10:41designed a a search window.
10:44And we defined a local 200 second
10:47window surrounding the end of the
10:50respiratory events as you see here.
10:52And so and and why we choose 102nd?
10:56Because usually it's 100 seconds
10:59longer than the longest events and and
11:02and that's why we use for a second.
11:04So you have this for this event you have
11:07this 202nd and then you go to the next event.
11:11And you do the same and you keep
11:14going until you cover all the
11:17events during the sleep.
11:19At the end,
11:21so we all the events were aligned based
11:24on the end of each event which is times 0.
11:28So we just put everything on
11:30top of each other.
11:32And did an ensemble averaging to
11:36get an idea of average behavior
11:40or average desaturation curve.
11:43For for each subject,
11:44so it's a subject specific search
11:46window that we used to calculate
11:49the hypothesis period.
11:50So once we have the search window,
11:53we go to each respective events
11:54and we have this time information
11:56from the search window.
11:58And then we calculate the area under
12:01the curve for each each event and then
12:05the hypoxic burden will be the total
12:08area divided by total sleep time.
12:11So there have been some
12:13similar metrics in the past.
12:15However, there are some key differences.
12:18For example, some of them were they
12:20were based on the desaturation.
12:23Ohh 24% and that this session
12:26desaturation were not linked to
12:29respiratory events and there were.
12:32They require visual detection
12:34of of incomplete recovery, so.
12:37So we tried to to resolve all
12:40all all these issues with with
12:43the development of 556.
12:45So this graph that you see here.
12:48On the X axis we have apnea hypopnea index
12:51and on the Y axis we have hypoxic burden.
12:54As you see there is substantial
12:57variability for any given HIV.
12:59For example,
13:00HIV of 30 hypoxia and go from
13:0550 to 150% minute per hour.
13:08So and hypoxic pattern of 50%
13:11minute per hour means.
13:13A 10 minute per hour of 5% desaturation.
13:17That's how we can interpret the values of.
13:22And here I'm showing the the
13:26ensemble average desaturation curve.
13:30For six different subjects,
13:32so you you see the HIV values.
13:36On the top and then hypoxic and
13:39values have just below below
13:41the HIV and you see the spike.
13:44The severity of desaturation.
13:46All these subjects have similar ages.
13:50Well,
13:50there's different patterns of the saturation,
13:53so it's it's important to.
13:58To try to better vectorize the
14:01their sporting events and and get
14:04the Disat area of of of the events
14:07and and measure the hypoxic event.
14:10Uh.
14:11Because it's it's better,
14:13it's a better metrics of of our
14:16system ability than just counting
14:17the number of events so.
14:21In this figure,
14:22I'm showing the T-90 on the X
14:25axis and hypoxic on the Y axis.
14:28So there are some subjects here.
14:33Uh,
14:34with a high T-90 but a low
14:38value of hypoxic period,
14:40these are these are the subsequent
14:43sustained hypoxemia as opposed to
14:46a related intermittent hypoxemia.
14:48And and the interesting part is
14:51that T-90 when T-90 is close to 0.
14:55There is a wide variation
14:56in the hypoxic burden.
14:58So for many,
14:59many subjects,
15:00the T-90 really doesn't capture
15:03the the intermittent hypoxia that
15:05we were captured with hypoxia.
15:10So we we have done several analysis
15:13to show the importance of hypoxic and
15:16the first was the association between
15:19hypoxic burden and CVD related mortality.
15:23We use 2. Observation observational
15:26cohorts Mr Oz and Stephen House Mr
15:31Oz older men average age of about
15:3376 years and they were followed for
15:37about 10 years and there were 440 CVD
15:40related deaths and the Sleep Heart
15:43House we have both men and women.
15:45And the H younger lady longer 64
15:49year old on average followed for
15:52about 11 years 313 CBT related death.
15:55And what we showed that so we there was
16:00a those are some response relationship
16:02with the severity of hypoxic burden and and.
16:07And increased risk of CVD mortality
16:09in these two cohorts.
16:14We then did some analysis with incident
16:17heart failure and as you see here,
16:19I'm comparing HIV versus hypoxic
16:21burden and you see a clear dose
16:25response relationship between if you
16:28use hypoxic burn compared to HIV.
16:31And this figure on the left,
16:34bottom left is is interesting.
16:37What we did here was a
16:39secondary exploratory analysis.
16:40We categorized individuals into
16:43low hypoxic burden and low HIV,
16:46which was our reference group.
16:49And then low hypoxic burn,
16:50higher HI, higher hypoxic burn
16:53and low HIV and high hypoxic IHI.
16:57And as you see here regardless
17:00of the level of HIV,
17:02high hypoxic burden were associated
17:05with increased risk was while if
17:09you have low hypoxylon and high,
17:12it's really there was no association
17:14with with instance or failure.
17:19And. So a study, a clinical cohort.
17:24In in that. From France,
17:28they actually tested hypoxic burden on.
17:31On our clinical cohort,
17:33I think it was HIV written five
17:35they included only and there were
17:38immediate follow up about 78 months.
17:435400 patients with OSA
17:46and without CD baseline.
17:48And they had a composite outcome
17:51of all cosmetology or CD.
17:54And they they saw a similar
17:57association with hypoxia and.
17:59And and the outcomes that
18:02they were investigating so.
18:05Oh, we did some cross-sectional
18:09analysis in Mesa study.
18:11And we looked at hypoxic burden and
18:15blood pressure and also hypoxic
18:18period and chronic kidney disease and.
18:22So these are the two studies
18:23that showed hypoxic pattern was
18:25associated with blood pressure,
18:26increased blood pressure and
18:27hypoxic term was.
18:28So it's associated with higher prevalence
18:31ratios of moderate to severe CKD.
18:37There have been other studies actually
18:41that looked at similar metrics.
18:44In the last two years and
18:47what they found was that.
18:49For example, this this metric which is
18:53called sleep breathing impairment index.
18:55Similar metric was associated with
18:58cardiovascular risk in male patients and.
19:03Also some association with daytime
19:06sleepiness better than HIV and
19:09and some neurocognitive outcomes.
19:11Here PVD based reaction time
19:14and severity of desaturation.
19:16So there have been some other
19:18studies that have been interested
19:21in Desaturation area under the
19:23curve and show some some significant
19:26association with these outcomes.
19:30OK, so now here I would like to talk
19:34about the terminal of hypoxic Berlin.
19:37So. Hypoxic burden in in sleep
19:42apnea is mostly related to reduced
19:44ventilation due to OSA. However.
19:49A recent research letter that
19:52was published in Blue Journal.
19:54They argued about the the
19:57contribution of abdominal obesity.
20:00To hypoxic burden.
20:03And and so and the lack of.
20:08Appropriate adjustment for abdominal
20:10obesity in these studies that we have done.
20:14So the argument was that this abdominal
20:17obesity leads to a reduced FRC,
20:21which leads to decreased baseline SP
20:24O2 and faster and deeper desaturation.
20:27Independent of ventilatory decrease,
20:30which is related to sleep apnea.
20:35So just trying to clarify more.
20:40Assume that we have two subjects.
20:43With similar OSA severity
20:45but different hypoxic burden.
20:48So what you see here is the ensemble
20:52average curve of. Ventilation.
20:57And the hi is 30 and 30 burden
21:00which is basically the area on the
21:03total area under ventilatory curve.
21:06During sleep per hour of sleep.
21:09And so this, this,
21:11this mentality burden leads to a
21:14hypothetical for this particular subject
21:18leads to hypoxic permanent about 83%
21:21minute per hour, as you see here.
21:23Assume that we have a.
21:26We have another subject
21:28with exactly the same age.
21:30The same age are the same entity Berlin,
21:33but with a lower hypoxic Berlin.
21:36So the argument that was done in.
21:40In in the this research letter
21:43was that this subject here in the
21:45top with higher hypoxic Berlin.
21:47Has a lower.
21:50Point volume, which results to more severe,
21:53deeper or faster desaturation. And.
21:58And and we really currently don't have.
22:03Unless we do some city scan of of the.
22:08Of the visceral fat,
22:10we really cannot adjust for this.
22:13And and and and the argument was that.
22:17This will affect the association
22:20between hypercycle and and and CBD.
22:24So we try to answer this question
22:27in this study that the manuscript
22:29is actually under review right now.
22:33So we we assess the relationship
22:37between hypoxic burden and ventilatory
22:39burden as well as available measure
22:42of abdominal obesity and other
22:44confounder in in Mesa sleep study.
22:46They had some astrometry parameter
22:49in Mesa long which we used to try to
22:53answer this question and we separately
22:57tested if ventilator burden predicted CVD.
23:01And we also did another analysis.
23:06We adjusted the association of
23:10HBCD for desaturation sensitivity.
23:13And and and so the.
23:17The the idea behind it,
23:18behind this is that,
23:21so this subject here in the top
23:24will have a higher desaturation
23:27or tendency to desaturate.
23:29So we try to to adjust for
23:32desaturation sensitivity which was
23:34which we defined as the amount of
23:36hypoxia that you get per ventilatory
23:39per reduction in ventilation.
23:41So we adjusted this for desaturation
23:44sensitivity to see if it actually.
23:47Anything changes and if these
23:49situations sensitivity by itself is
23:52actually predictive of of the outcomes.
23:55So we did the study in Mesa sleep.
24:00Study about 1950 subjects so the
24:03definition of mentality burden
24:06which he defined it as event
24:09specific area under ventilation,
24:11ventilation signal so mean
24:13normalize and air we calculate the
24:16area under the mean ventilation.
24:19And the outcomes that were tested
24:22was instant cardiovascular disease,
24:24instant coronary heart heart
24:26disease and all cosmos totality.
24:30So this is briefly shows how
24:33we measure penalty Berlin.
24:35So if you look at the ensemble
24:38averaging of the ventilatory curve,
24:41so we can get the average event duration,
24:43we can get the event depth.
24:46And events to burden can simply
24:49be defined as the event rate times
24:52vent depths time event duration,
24:54which is a measure of total total
24:58ventilatory deficit during sleep.
25:01So this table shows the baseline
25:05characteristics of the Mesa
25:07sleep study that we included.
25:11So the average age about 67 year and.
25:18About 5053% of women.
25:23Equal distribution of different
25:25arrays and necessities and you see
25:29the PMI here. What we have here?
25:31Is that we have the HIV,
25:34the median HIV in this cohort was
25:37about 33 events per hour when
25:39switching Berlin and hypoxic Berlin.
25:41So in Mesa as just when I get a feel
25:45of how much rental through loss
25:47there was on average in this in
25:51this population cohort was about 20%
25:54was about 3 minutes of apnea which
25:57is the 100% reduction in airflow
26:00per hour of sleep and the hypoxic.
26:03But 9 minutes of 4% desaturation
26:05per hour of sleep.
26:09And this graph shows the association
26:11between ventilated burden and hypoxemia.
26:14You 17 burden on the X axis and
26:17hypoxic burden on the Y axis.
26:19They were strongly correlated
26:21and the R score was about 0.8.
26:25In this table I'm showing the
26:29contribution of other factors.
26:31So the Model 1 included mentality burden.
26:36BRS score of 0.8 in Model 2 we added a BMI.
26:42And and body surface area.
26:44So the R-squared increased only by 1%.
26:48Model 3 as you see here,
26:51we added wakefulness to baseline SPO
26:542 because there was one argument
26:56that baseline SP O2,
26:58the lower the baseline the the
27:01deeper the saturation and that could,
27:04independent of ventilatory deficit,
27:05affect the values of hypoxic.
27:08And so we added. So what about 1% additional?
27:14Variation explained by adding
27:16these two variables here.
27:19And and and similarly in the other models
27:22where we did some spirometry parameters,
27:25so there was no change in the R square.
27:29So and and what what it tells us is that the.
27:35Did the variation in hypoxic furnace
27:38mostly described by reality burn
27:40which is the OC related component
27:43that we are interested in?
27:45So what you see here is the hypoxic
27:48pattern and venture burden association
27:50with instant CD and all Cosmo thority.
27:54So you have a instant CHD and this
27:56is a hazard ratio for hypoxia.
27:59Incident, CVD and all cosmetology.
28:02And you see a similar association
28:05with Renton Superman.
28:07And these these these hazard ratios were
28:10adjusted for age 6 phase BMI hypertension
28:14as well as the desaturation sensitivity.
28:17So we also adjusted for desaturation
28:21sensitivity to to adjust out those
28:24other unseen or unobserved confounders.
28:28That we really cannot measure in
28:33this large population towards.
28:36So the take home message is that
28:39hypothesis is minimally affected by
28:41available measure of either positive
28:43and lung volume and ERV do not vary
28:46much in OSA population the action in
28:48fact we looked at the coefficient
28:51of variation of of these different
28:54factors waste circumference the
28:56coefficient of variation was about 14%,
28:59BMI 19% while for rental to burden
29:02is was about 100% of the coefficient
29:06of variation. And and the.
29:08Another take home message is the
29:10hybrid and largely captures the risk
29:13attributable to Ben 30 burden of OC
29:16rather than the tendency to desaturate.
29:19And we favor hypoxic burden over
29:22ventilated burden because ventilation
29:24is usually more difficult to measure.
29:26In a home based setting and their
29:29calibration issues and lack
29:31of standardization.
29:34OK. So that was related to hypoxic planets.
29:37Next I moved to Hartford response
29:39to Afghans and Hypopneas.
29:41So this this is slide was actually.
29:46From doctor summers. Paper in 1995.
29:51Where he described sleep apnea and
29:54an increase in sympathetic activity
29:56and and increasing blood pressure.
29:59As you see here with each respiratory events
30:03there are an increase in in heart rate.
30:06So we were thinking so if by
30:09measuring this this delta heart rate,
30:11increasing heart rate with each event we
30:14may gain additional information that and
30:18that may be useful for risk prediction
30:21or for to to identify who respond to
30:25CPAP treatment and who benefit from C.
30:28So some patients have a larger heart
30:31rate response to events than others.
30:34Here I'm showing two different subjects.
30:38And in the bottom figure here you see the
30:41increase in heart rate by this Red Arrows.
30:45So this subject has a minimal increase
30:47in heart rate with each respective event,
30:49despite similar oxygen desaturation amounts.
30:54And while this subject has a
30:57larger increase in in heart rate.
31:00So Delta Hartrick can be easily measured
31:02from in in lab and in home PSG's and
31:05currently is really under utilized.
31:07So and we our previous study back
31:11in 2013 we showed that delta heart
31:15rate increases with event severity.
31:18And and in observations without
31:20cortical arousal and with arousal.
31:22So there was a similar pattern of increase.
31:25Obviously if you have a razor,
31:27you have the biggest,
31:29the the largest increase.
31:31So the hypothesis was that high delta
31:35heart rates is associated with increased
31:38risk of cardiovascular disease.
31:40And we separately hypothesize that
31:44removing respiratory stimuli.
31:46And and those are apneas and hypopneas
31:49by CPAP.
31:50Could be beneficial in in those
31:52with High Delta heart rates.
31:55So to test the first hypothesis,
31:57we looked at the Mesa and
32:00Steve Hartman study.
32:01And and this here I'm showing
32:04this sample flow chart.
32:06At the end about 14 hundreds were
32:10analyzing Masa and about 4600
32:13analyzing from Cpot health study.
32:16So we saw a U-shaped association
32:19between Dalton Hartry and subclinical
32:22CVD measures including.
32:26Calcium score NT,
32:27probie NP and Framingham CVD risk.
32:31And based on this,
32:32we categorize the doctor heart rate in low,
32:35mid and high groups.
32:36And we looked at, we tested that.
32:40Categories and and sleep harthouse.
32:43And what we saw was that indeed
32:45those with low and High Delta Party,
32:48they were at increased risk of
32:51CVD and nonfatal CVD,
32:54CVD mortality and all 'cause mortality.
32:57So, so basically.
33:01The risk observing the low delta
33:03heart rate group we we thought and we
33:06hypothesized that was actually non OC
33:09specific and mainly due to heart disease,
33:12diabetes or autonomic dysfunction and and
33:14the risk observing the height of the group.
33:17Was mostly OSA specific because we
33:21saw hired hazard ratio in those with
33:24severe OSA whether quantified by HIV
33:27or hypoxic German and the risk was
33:31exclusively observing those with ESS
33:33less than 11, so non sleepy OSA.
33:36So the next question was.
33:40If with CPAP, reduce the risk in
33:42those with high throughput rates.
33:44And and to answer this questions we
33:47we looked at the recuts of trial.
33:51So request the trial was done and
33:53it was published actually in 2016,
33:56non sleepy patients with OSA
33:58and coronary artery disease
34:01randomized to CPAP or no CPAP. And.
34:05The HI criteria for inclusion was
34:08greater than 15 ESS, less than 10,
34:11median follow-up was about
34:1357 months and the outcome?
34:17Was a composite of first event
34:20of repeat vascularization,
34:21my MI stroke or cardiovascular mortality.
34:25And we asked the question is the CPAP
34:28benefit contingent on the heart rate
34:30response to events and what we saw is
34:33these graph here show the point estimates.
34:36At average delta heart rate 6 beats per
34:40minute and this is what's recuts trial.
34:44The original recursive trial showed
34:46was about 0.8 non significant
34:49and low delta heart rate.
34:51There was a suggestion of harm and
34:54so treating people with no low
34:57Delta Harvard seem to I was non
34:59significant and very wide confidence
35:01interval and but within those with
35:03high tops of heart rate,
35:05higher delta heart rate.
35:07We saw a risk reduction at
35:10significant risk reduction.
35:14Compared to other group so.
35:17And and the model were just for age 6,
35:20PMI and cardiac intervention.
35:25This is basically showing the same thing,
35:27but we have binary subgroup analysis.
35:30We have all participants gained similar to.
35:35That's a trial, so it was no effect.
35:38If you look at Delta heart rate
35:41threatens it's it's per minute you
35:43see that the CPAP significantly
35:46reduces the risk compared to this.
35:51So in summary, a greater Harter
35:53response to respiratory events is
35:56a risk factor that's identifiable,
35:58deleterious, and potentially reversible.
36:00And it could be used to select patients
36:04most likely to exhibit long term
36:07cardiovascular benefit from CPAP treatments.
36:09I think I have time.
36:13Do I have maybe 6-7 minutes
36:15to talk about this or?
36:20Yep. OK. Yeah, that sounds good.
36:23Thank you. So in and so this this work
36:27is currently under review and we're here
36:31we we're interested in the relevance of
36:34EEG arousal for risk stratification.
36:37So really the question was,
36:38do arousal improve prediction of OC
36:42related outcomes on top of desaturation?
36:44So over the past four decades, several
36:47versions of OC definition have emerged,
36:50which led to confusion among patient,
36:53clinician and payers.
36:54And and you were interested which
36:56type of hypo you know went to the
36:59desaturation of arousal would inform
37:00research stratification and the reason
37:02to do this study that was this ASM
37:05funded study that to systematically
37:07compare events with arousal and
37:09the saturation in multiple cohorts.
37:11So we used sleep quite health
37:13Mesa and slots for this study.
37:16So we compared events with
37:18minimal desaturation and arousal.
37:21So that's called HIV arousal only.
37:25And events with the saturation
37:27and no arousal.
37:28That's called H greater than 3% only.
37:32So these are the two.
37:35To specific HIV that we were interested in.
37:38So just quickly showed that sample
37:42characteristics of sleep heart
37:44health have shown that in the
37:47previous slide Mesa and Mr Oz and.
37:49The outcomes that we were interested in
37:52was so there were some cross-sectional
37:55outcomes including hypertension,
37:56diabetes and sleepiness and some
37:58follow up data longitudinal data
38:01instance CD in in these three cohorts.
38:04So again so this table shows
38:08the total number of events.
38:11Analyzed in these three cohorts.
38:17And the distribution of events with more
38:20than 3% desaturation only about 54%.
38:2543% in Mesa and in Mr Ross for the 43%.
38:29And but 13% of events would with
38:33arousal only and no desaturation here,
38:37but you see here.
38:38These are the headlines of those events.
38:41And the so this translate to this
38:44HIV values that you see here.
38:47So if you if you include these
38:50events events with arousal only.
38:56So prevalence of Margaret Zero
38:58say would be 67 percent, 70%,
39:01sixty 8% in this state cohorts.
39:04But if you exclude these events,
39:06you get a 10% reduction from
39:10supposedly across street cops.
39:14So it's important to really look at these
39:16and see if there is any any additional
39:19information that these is provided.
39:21So what what you see here is the.
39:26Additional role of events with this
39:28saturation, but no other else.
39:31And the model were adjusted for covariates
39:34and HIV based on events with arousal.
39:37So regardless of desaturation.
39:39As you see here, across all all the
39:43outcomes you see as significant.
39:46Increase in in in risk.
39:48But if you look at the HIV arousal only.
39:53There's basically nothing
39:55so additional those events.
39:58If you adjust for desaturation events
40:00with these saturation of the also Brazil.
40:03So this what what this is is basically.
40:07Assume that you send the subject
40:10and you send an HD to the subject.
40:13And you get all the 3% events
40:16with more than 3%,
40:17so you don't measure arousals.
40:20And then you bring them back to the sleep
40:23lab and you find those additional events.
40:26That were associated with arousals.
40:29And what what we see here that
40:32those really don't add any info,
40:34don't improve the prediction of outcomes.
40:38As you see here,
40:39and in some cases they're even protected.
40:43So this I think we didn't adjust
40:46this for BMI so this this.
40:48This table which is BMI again you see
40:53similar direction and and slightly.
40:55A lower heat hazard or odds ratio,
40:59but it's similar story that we had before.
41:05So what we did this was an
41:08additional exploratory analysis.
41:09We looked at individuals now with
41:12an HI getting 15 events per hour.
41:18Again, based on H 3% regards of arousal.
41:22So exclude everyone who had
41:25severe OSA by desaturation.
41:28Audit to Studio City and again in
41:30this in these individual events with
41:33arousal put but no desaturation.
41:36Were not associated to it with this stuff,
41:38adverse outcomes. So similar,
41:40similar similar results that we've found.
41:45And and to further investigate this,
41:48we looked at the arousal
41:50intensity and arousal index.
41:52I think this was just be part of the study.
41:55You see, that's the arousal.
41:56Intensity tends to go down as well if
41:59you get more arousals per hour of sleep.
42:03So there is a negative association here
42:05and we did another analysis, we looked at.
42:09Change in arousal index and the
42:12sleep heart health and what we saw
42:16was that changing arousal index was.
42:20Predicted by baseline, Oasis, Verity what?
42:23What it tells us is that over time there
42:26is a decline in a number of arousal.
42:30And separately in another study
42:35that my poster Gonzalo Labarca did.
42:39We looked at a razor burn.
42:42Which we defined.
42:43It is similar to the.
42:45Paper published by the Australian
42:48group event rate times arousal duration
42:50and we compared it with hypoxic burden.
42:53So again. Compared to hypoxic burden,
42:56there is no significant
42:58sociation with arousal burden.
43:03So to take your message is
43:05while arousal may be harmful.
43:07Based on experimental data,
43:09measurements of arousal may not
43:11improve its positions department.
43:15So in summary, measuring the depth
43:17and duration of this saturation.
43:19That should provide added predictive value.
43:23So we can measure hypoxic from there than
43:26from inlab and in home sleep studies.
43:29We demonstrated that hypothesis paramedics
43:32multiple outcomes in population base
43:35and and French group in Congo cohorts.
43:39We show that hypoxic burden is strongly
43:42associated with entity burden.
43:43And is minimally affected
43:46by the abdominal obesity.
43:48And we also showed that individual without
43:52sleepiness the elevated heart responsiveness.
43:57Those will be at increased risk
43:59of CVD and may benefit from CPAP.
44:02And additional advanced with
44:03arousal and all the saturation to
44:06not improve this prediction.
44:08And finally,
44:09I'd like to thank you the amazing
44:12team in Boston and elsewhere for.
44:15Great contributions to to help with the,
44:19the data,
44:20with the analysis and and and and everything.
44:23And thanks everyone for.
44:26Listening to my talk.
44:28Thank you so much. That was fantastic,
44:31very informative and really enjoyed
44:34hearing that the different algorithms
44:36you've helped develop or developed
44:38yourself and how they've been
44:40applied to different clinical trials.
44:43I'm going to take questions now,
44:44but I'll ask if you could please
44:47put your question in the chat and
44:50let me call on you rather than
44:52unmuting and and so we can avoid
44:55the risk of missing questions or
44:58not giving everyone a chance. Umm.
45:01So does anyone have any questions?
45:08Doctor Thomas is asking is loop
45:10gain elevated in those with
45:12a high heart rate response?
45:16We looked at this, I think the
45:18correlation between delta heart rate
45:21and Lucan was about 0.2 positive.
45:27Yeah. So it's just like, right, ability.
45:31And then Doctor Winkelman says it looked
45:33as if low hypoxic burden and high age chi
45:37were protective from mortality comment.
45:41I participated and high.
45:44Low hypoxic burden, high I
45:45had lower risk of mortality,
45:47yes, it was actually, it was insulin
45:49heart failure that we found, yes. Yeah.
45:53So lower risk of incident and
45:55heart failure with those with
45:56a high and low hypoxic burden.
45:58Yeah, it wasn't significant.
46:00So yeah. The hazard ratio I think
46:04was 0.9 or something like that.
46:08And clinically that makes sense because
46:10with heart failure patients in particular,
46:12they can often have very long apneas and
46:15in combination with the circulatory delay
46:18it can be much more detrimental than
46:21than a patient without heart failure.
46:23Doctor Thomas says patients
46:24come with now symptoms,
46:26not for future prevention mostly,
46:28so arousals should not be ignored.
46:33Yes, this has been a very controversial.
46:37Study that we, I mean we didn't
46:39expect that to see this results.
46:42But and and I mean it may be
46:44in in younger individuals if.
46:50The beginning of the disease.
46:52They may be very informative, but.
46:55These, these are mostly older individual
46:58and and I think there was another there.
47:01There was a study that showed the
47:04initiation of symptoms and that
47:06the time between the initiation of
47:09symptoms and the official diagnosis
47:11whatever was about maybe 11 years.
47:14So by that time the arousal may
47:16not be informative any anymore.
47:18So and in terms of risk prediction.
47:22So that's that.
47:23I need to make that clear
47:25so it doesn't predict.
47:27Addition the added risk and top of
47:30desaturation based on these studies.
47:36I'm going to go back to Doctor
47:37Hoffman, doctor Yaggi said.
47:38Do we know the relationship
47:40between arousal threshold and
47:41delta heart rate response?
47:45No, I haven't done that study.
47:48So arousal threshold and delta heart rate,
47:51no, I don't. I don't know.
47:53But yeah, we can, we can look at it.
47:57And then, doctor Huff. Oh, go ahead. Sorry.
47:59No, I was just curious why.
48:02Like the gag is interested in this.
48:05Please unmute and and let us
48:07know Doctor Yaggi if you can.
48:10OK.
48:13And then Doctor Hoffman, if you want to
48:15unmute to ask or to make your comment,
48:18I bring that up just because the
48:21work that Andre has done has
48:24shown that low arousal threshold,
48:26the patients really struggle adhering
48:29to CPAP therapy and we're finding that.
48:34The arousal threshold
48:35and autonomic activity might be
48:38related that it might be challenging
48:40to treat those patients with delta
48:42heart rate response if they if
48:45it's an arousal threshold issue.
48:48That's interesting. Sure. Yeah,
48:50we should look at it. We have the data.
48:55Doctor Hoffman wanted to point out
48:56that using heart rate response can be
48:58tricky when we don't know if the patient
49:01has a cardiac autonomic neuropathy,
49:02which is not uncommon in patients with type
49:052 diabetes and and as patients are aging.
49:11I don't know if you have a comment for that.
49:15I mean I don't have a comments, but I agree.
49:18Yes, yeah, in in some patient
49:20population it may be challenging
49:22to measure heart rate response,
49:24but we haven't had any issue
49:25in in this large cohorts,
49:27population based cohorts.
49:29In terms of measurements of delta,
49:31heart rates and.
49:34And Doctor Winkleman, do you
49:35wanna unmute for your question?
49:37It's a little bit longer if you'd like.
49:46Thank you. Great talk, Ollie.
49:49Progression of.
49:51A very logical approach to this.
49:55I think from my perspective
49:58the next step is looking at
50:01respiratory related leg movements.
50:04Those leg movements that occur
50:05at the end of respiratory events
50:08that have been shown when they
50:09are present to be associated with
50:12a substantially larger increase
50:13in heart rate than events that do
50:16not terminate in the leg movement.
50:19Do you have any plans to investigate this?
50:23Absolutely. I mean, we are doing that.
50:27That's a good time to ask this question.
50:31We're trying and trying to
50:33have some data. To analyze that data
50:37and hopefully, yeah, collaborates.
50:40Thank you. Great talk. Thank you.
50:44And uh, Doctor Thomas ask,
50:47how does one use these insights in
50:49the management of individual patients?
50:51So you've shown a group analysis with
50:54with large cohorts of patients and do
50:56you have any recommendations to us to
50:59tailor this clinically at this point?
51:01I don't, I don't. Yeah.
51:03That's that's too, too early.
51:05Probably it's too early.
51:06Yeah. Yeah. Yeah.
51:09Um. And then someone is asking arousal
51:12response is related to obstructive
51:14sleep apnea syndrome or neuropathy?
51:19Uh, if you want to unmute, it's CAI.
51:22If you want to unmute and and.
51:26Clarify the question would be great.
51:43Is that Alice Kai? I'm not sure.
51:48Alice, you want to unmute and.
51:51Alright. Yes, I just want to know
51:55um arouse response is result
51:57is related to the authors or
52:01maybe from the central obstacles
52:05central right you're passing?
52:09Central sleep apnea.
52:16Can you clarify again? Sorry,
52:17we had a little difficulty hearing you.
52:21Means uh arouse response is related
52:26to obstruct sleep as clear syndrome
52:30or symptoms sleep aspirator syndrome.
52:33We only looked at the obstructive sleep
52:37apnea. We didn't get this central.
52:42And their central events were very
52:44rare actually in these these cohorts.
52:50Doctor Yaggi, I'm going to ask you
52:51to any new again as well to share
52:53your comment if you don't mind.
52:57Yeah. One one other question.
52:58I'll leave this unbelievable work.
53:00Thank you for sharing this with us today.
53:03It seemed like in the analysis
53:05you did looking at Delta heart
53:07rate in the observational studies,
53:09I think it was sleep heart health,
53:11Mr Ross Mesa where it's at a low,
53:14medium, high heart rate cut points
53:16where you dichotomize this and the
53:19CPAP analysis for ricotta using
53:22different different cut points.
53:25And I'm just wondering.
53:26The reason behind that was that yeah,
53:29how did you come up with those cut points
53:31and ricotta versus versus the other
53:34side. So sure. Thank you for comments.
53:38So basically in in sleep in Mesa
53:41and sleep Heart health actually
53:43in Mesa we looked at this U-shaped
53:47relationship and based on that.
53:49We just categorize based on the
53:52top 25% and and lower 25% and
53:57the middle was the middle 50%.
54:00And and then we use the same
54:02threshold in sleep Charter.
54:03So we didn't change the
54:05threshold in sleep harthouse.
54:06In regards to the,
54:08the question was different actually
54:10it's because we hypothesize that
54:13those with higher heart rate
54:15response those would benefit more
54:17and and we test it as a linear.
54:21So we test it as a.
54:22We didn't in the main analysis we
54:24it was a continuous variable and
54:27we showed that there was that,
54:28I showed the point estimates and the.
54:31The the binary.
54:33Categorization that was just just to
54:35show that you see the same thing if
54:38you divide it by a nice around number,
54:41like 6 beats per minute.
54:44And and and there is also a
54:45difference in the in the population.
54:47So in the recuts I think it was
54:4990% or maybe 80% had cardiac,
54:54cardiac problems and about
54:5690% were on beta blockers so.
54:59It would be a different population compared
55:01to State Park health and other records.
55:08Sorry, I couldn't hear the rest.
55:11Or Andre, feel free to unmute,
55:12unmute if you wanna.
55:14No sure. So I'll, I'll leave.
55:16Great, great work and like Doctor
55:19Wickman said are really nice and
55:22elegant progression of going from
55:23HIV to these more continuous metrics,
55:25which is really nice.
55:26And I guess I have a couple of questions.
55:29One question is?
55:30You know, I think that, Umm,
55:32the markers you've developed
55:33are very promising and I'm just
55:35wondering if you were to take it
55:37to the next level or someone else
55:38were to take it to the next level.
55:40You know what,
55:41I'm going to sort of piggyback
55:44on Clara's question.
55:45You know, how do we design a trial
55:47where we look at people most at risk
55:50or most likely to respond for example.
55:52So what delta heart rate caught up do we use?
55:55Do you think we need to validate
55:57this more in other populations
55:58before you have a cut off?
56:00How can we operationalize that to
56:02the point where we can perform a
56:04trial that's a little bit more?
56:06Umm.
56:08Successful.
56:11Yes, we are trying to to do that trial
56:14and submit the proposal and hopefully
56:17get that fund that's for the delta
56:20heart rates and we proposed that delta
56:23heart rate of 6 beats per minute. OK.
56:27Because again we're going to recruit
56:30from cardiology clinic most likely.
56:32And yes, that's I'm going to use the same,
56:36the same threshold that that was in bukasa.
56:41You know. So.
56:44OK. All right.
56:47Three of six cardiac population.
56:48Here we go. All right, excellent.
56:50And I guess the other question I had was?
56:54In regards to your more recent work
56:56where you're looking at the risk
56:59of adverse outcomes and different
57:01definitions of the HIV that we could
57:03extract from the routine PhD data
57:05without having to use signal processing.
57:08And so when you're looking at events that
57:10are associated only with arousals versus,
57:13those are the 3% disabled.
57:15So in that population,
57:16what was the distribution of the
57:19events that were associated with 3%
57:22dsat versus events with just the?
57:24There's always is arousals alone,
57:27minority of events or whether
57:28I think I showed that in
57:30the one of the tables,
57:31I think yes, about 15% or so,
57:37yes, only arousal notice,
57:40right? So in your analysis you
57:44essentially you add her on the
57:46arousal frequency really you know
57:49register event related arousals or
57:52you know hoping really arousals.
57:54The total amount of the exposure
57:56that's already there with the 3%
57:59exactly and that's exactly what we did, yes.
58:03All right. And one of the other
58:05things I might suggest is that,
58:06you know, as we look at these
58:08different different studies,
58:09one of the key pieces of information is
58:12that often we use composite outcomes,
58:14you know, CAD stroke and they actually might
58:17be different mechanisms for those events.
58:20And similarly, you know, maybe the
58:22arousals may be more relevant for you know,
58:25cognition or rather than CVD.
58:28And so there's some data I think
58:30from Stanford looking at the.
58:33Hypopneas with arousals only would
58:35be correlating this objectives.
58:39I mean, we looked at the ESS,
58:40but you're right, yes.
58:42So we should look at other
58:44outcomes that are available,
58:46I think, yeah. Unfortunately,
58:49in this course is I think ES was
58:52the only one we could look at.
58:57Our next question asks what clinical
58:59parameters would help delineate
59:01individuals more likely to have
59:03significant hypoxic burden when
59:05they're being evaluated and in.
59:08So. So the. In terms of the,
59:13so you mean people with maybe high HIV,
59:16high Tina, is that the question?
59:20Please feel free to unmute
59:21the individual asking this.
59:23I think they mean.
59:24I'm wondering if they mean in the clinic.
59:25Are there any characteristics we can
59:28use to to sort of identify those?
59:31But I could be wrong.
59:36That's from
59:39EN 12901.
59:42I mean if you see a lot of events.
59:45A lot of these saturation,
59:47I mean in that meeting before before
59:50actually the study like you in the clinic
59:52you evaluating a you know patient.
59:56You know I understand that you know all
59:58that is from the study but you
01:00:00know clinically if we evaluating
01:00:02patients since they are you know pack
01:00:05some car clinic or characteristics that
01:00:08you know we could. Basically,
01:00:11I derive from the studies
01:00:13that would predict that those
01:00:14individuals are more likely to
01:00:18have significant hypoxic burden.
01:00:20Done that that study.
01:00:25You showed that abdominal adiposity
01:00:27wasn't associated with it,
01:00:29but were there any other anything like?
01:00:32Any characteristics?
01:00:36To some extent. Uh.
01:00:39Which is related to both sleep
01:00:41apnea and also maybe just
01:00:44adding some more hypoxia.
01:00:46By lowering the lung volume.
01:00:50I think you have BMI was the only
01:00:52factor that remained significant after
01:00:54it's just include the covariance.
01:01:00Ali, thank you so much.
01:01:02What a great talk and thank you, so good
01:01:05for hosting a lot of outstanding data,
01:01:08a lot of promising directions and
01:01:11looking forward to hearing from you in
01:01:13a couple of years when you update us
01:01:16and tell us that you figured it out.
01:01:18And looking forward to our next
01:01:20session that we presented by our
01:01:22own doctor minor at Yale,
01:01:24we'll be talking about sleep
01:01:26disturbance in the elderly population,
01:01:28something that we see fairly commonly.
01:01:30And so thank you everyone for
01:01:31spending your time with us today
01:01:33and we'll see you in one month.
01:01:35Thank you so much for hosting.
01:01:37Bye, bye. Thank you. Bye.