Can Machine Learning Predict CVD Risk? Sleep-Related Benefits of Exercise

TTHealthWatch is a weekly podcast from Texas Tech. In it, Elizabeth Tracey, director of electronic media for Johns Hopkins Medicine in Baltimore, and Rick Lange, MD, president of the Texas Tech University Health Sciences Center in El Paso, look at the top medical stories of the week.

This week’s topics include deep learning for predicting cardiovascular events, molnupiravir in immunocompromised people with COVID, sleep and exercise, and oral steroids in preschool wheeze.

Program notes:

0:42 Molnupiravir, COVID and immunocompromised people

1:42 Samples up to 44 days post treatment

2:42 Causes mutagenesis

3:44 Ongoing surveillance?

4:14 Oral steroids in preschool wheeze

5:21 Greater reduction in wheezing severity score

6:21 How often it recurs?

6:33 Use of machine learning to estimate CVD risk with radiographs

7:35 Compared with 2,000+ with known risk

8:35 Routine chest x-ray informs risk

9:35 Exercise and sleep quality

10:35 Physically active at beginning and 10 years later

12:12 End

Transcript:

Elizabeth: Can deep learning help us to predict cardiovascular events?

Rick: Should you give steroids to young kids that wheeze?

Elizabeth: Molnupiravir in immunocompromised people with COVID-19.

Rick: And do people that exercise a lot sleep it off?

Elizabeth: That’s what we’re talking about this week on TTHealthWatch, your weekly look at the medical headlines from Texas Tech University Health Sciences Center in El Paso. I’m Elizabeth Tracey, a Baltimore-based medical journalist.

Rick: And I’m Rick Lange, president of Texas Tech University Health Sciences Center in El Paso, where I’m also Dean of the Paul L. Foster School of Medicine.

Elizabeth: In keeping with our long-standing policy, let’s turn first to The Lancet. This is a look at something that we have all been talking about and suspecting ever since COVID emerged — oh, my goodness — 4 years ago. This is a look at molnupiravir, so the antiviral that’s frequently given to people with COVID-19 infection, given to folks who are immunocompromised and then a look at what happens to that virus in those people.

The hypothesis that the authors start with is that continued SARS-CoV2 infection among immunocompromised individuals is likely to play a role in generating genomic diversity and emergence of novel variants. This, of course, was one of the nightmare scenarios that we all considered when COVID-19 was really flying around the planet.

This is a very small study. It includes five immunocompromised patients who were treated with molnupiravir and four patients not treated with it, two of whom were immunocompromised and two were not. They collected — and these folks, we gotta give them kudos — throat and nasopharyngeal samples in these patients up to 44 days post-treatment and then they sequenced the viruses.

They followed that by what’s called “variant calling” and what they find is molnupiravir showed a large increase in low-to-mid frequency variants in as little as 10 days after treatment in those folks who were treated. The untreated people didn’t see this and some of these variants became fixed in their viral population.

Molnupiravir treatment, then, in immunocompromised patients led to an accumulation of a distinct pattern of mutations beyond the recommended 5 days of treatment. These folks maintained a persistent PCR positivity for the duration of their monitoring that may increase their potential for transmission to others and the subsequent emergence of novel variants.

Rick: This particular drug, molnupiravir, the way it works in COVID is it actually causes mutagenesis. The fact that it mutates it so that it’s no longer effective is how it helps clear it. Now, it’s not a first-line drug; it’s actually a third-line therapy. But in people that are immunocompetent, because they clear the virus after 5 days, this mutagenesis or the change in COVID probably isn’t a big deal. But it is in immunocompromised individuals because they have persistent COVID infection.

If you cause a mutation in the COVID virus, it can persist. In these individuals, it persisted up to 44 days. It just happens to happen rarely in a large number of individuals before it’s a real problem.

Elizabeth: It’s unclear to me, however, what do we do about people who are immunocompromised and become infected with COVID or with other viruses. I bet this is not the only virus that this happens with.

Rick: This is not first-line therapy. We have ritonavir-boosted therapies, remdesivir, and they are recommended as first-line therapies. Molnupiravir is only recommended as an alternative therapy when either the other ones aren’t available or for some reason is clinically appropriate.

Elizabeth: Would you say that ongoing surveillance of people who are immunocompromised would be prudent?

Rick: In fact, that’s one of the recommendations. This would be very difficult to have the vigilance to repetitively test these individuals over the course of weeks and months.

Elizabeth: I wonder about self-isolation for them.

Rick: Ideally, if someone has persistent COVID infection, you would isolate. That would be the recommendation.

Elizabeth: Okay. Which of your two would you like to turn to?

Rick: Let’s talk about the use of oral steroids in what I’m going to call “acute preschool wheeze.” Studies have shown that steroids are beneficial in kids that have asthma, and that’s usually kids that are over 6 years of age, and we have known that steroids are not beneficial in kids that are wheezing less than 1 year of age. But we got this middle group of kids between the ages of 1 and 6 — when they wheeze, should we be just giving them steroids?

There is some controversy. There were some groups that recommend, “Yes, you should give steroids to kids between the ages of 1 and 6;” others that don’t. As a result, if you look at the practices across the U.S., they are administered in as few as 20% of kids or as many as 80% of kids in this range.

What these investigators did is they tried to summarize and look at all of the data, and looked at individual participant data so they could do a meta-analysis. These are trials published from 1994 to 2020 with kids aged, again, 1 to 5. They had preschool wheeze and they looked at the wheezing severity. Also, if they were hospitalized, how long were they in the hospital? There were 2100 kids, some that had received corticosteroids and some that had not.

What they determined was, those that received steroid treatment had a greater change in their wheezing severity score — it actually went down at 4 hours — and there was a decrease in the hospital stay for those that were hospitalized. What’s the downside of it? Because this was an acute, and not chronic, administration of steroids, there really wasn’t much in the way of downsides. There was an increased risk of vomiting with the kids. That was really about it.

The other thing that the study showed was it was effective in kids that had moderate to severe wheezing, not with mild wheezing, and it’s better the earlier you administer it. When the kids hit the emergency department, that’s when you start

Elizabeth: How does this fit in with all the rest of the chronic management strategies for kids with asthma?

Rick: This is different. They didn’t establish what the cause was in these particular kids. You think, “Well, gosh, if this was asthma, then kids that had a higher history of allergic reactions would benefit more.” But that wasn’t the case here. Some of this maybe had been viral. Some of it may have been asthma.

Elizabeth: I guess I’d like to look at longer term data relative to this population to0 — how often does it recur, how often would it mean if you use them once, and are you going to have to use them more often in the future.

Rick: Yep. In that particular case, that would more likely be asthma than any other cause.

Elizabeth: Let’s turn to Annals of Internal Medicine. This is taking a look at deep learning, machine learning, whatever you want to call it, to estimate cardiovascular risk. In this case, they’re using chest radiographs, which are extremely common, routinely done, and then we use them for more than one purpose. That’s what it sounds like to me.

As we know, there are these guidelines for primary prevention of atherosclerotic cardiovascular disease and they recommend a risk calculator that estimates a 10-year risk for major adverse cardiovascular events. Basically, this study is saying, “Gosh, if we train our computer algorithms to take a look at routine chest radiographs, what does that add to that 10-year risk score?”

They externally validated their model — and by the way, these folks have done some modeling, so they have got a lot of experience with it — in just shy of 9,000 outpatients with unknown cardiovascular risks because they had missing data inputs, and then they looked at 2,132 with known risks where that risk score could be calculated. Then they took a look at this additive value.

They found, sure enough, that taking a look with their model at a routine chest X-ray was able to accrete in a positive way to that 10-year risk score and may end up helping to identify individuals at high risk who might have missing data from other data sources.

Rick: Well, Elizabeth, I approached this particular study with a fair amount of skepticism. We know that we can estimate risk. What’s your risk over the next 10 years of having some cardiac event based upon things like your age, your gender, your cholesterol, the presence of hypertension, do you have diabetes, do you exercise, if you have a family history? All those things contribute.

Those are the things that we routinely use. We have different models we can plug in and get pretty precise numbers, and they’ll direct our therapy. But as you mentioned, there are individuals that we don’t have all this data, but, for example, they have a routine chest X-ray for something. Okay. If we don’t have all that data, get a routine chest X-ray, give us some information that puts people in a higher risk, and that would be helpful to know because we’d intensify therapy.

A couple of things about this particular study. One is when you do this deep neural networks it’s actually a black box. They don’t know how the computer decided. As a result of that, it’s very difficult to explain on an individual basis. They are not recommending that people get a routine chest X-ray. They are saying, “If you already have a chest X-ray, can we use it?” Furthermore, they wouldn’t treat anybody based upon it. But they say, “Hey, this person may have high-risk features. If you don’t have that other information that I mentioned, maybe you ought to get it because it may change their care.”

Elizabeth: I would suggest to you that in our emerging, more complete medical records for many people, I bet this is the kind of data that is going to be available on a lot more people.

Rick: It will be. We’re going to be using artificial intelligence and neural networks in the future and we’re going to be seeing many, many more studies like this.

Elizabeth: Then finally, let’s turn to the BMJ.

Rick: Do people that exercise sleep it off? We know that exercise is associated, really, with many things that improve health. You can lower your body mass index. You can lower your blood pressure. You can lower your cholesterol. It lowers your cardiovascular risk.

But there are very few studies that have actually looked at physical activity over a long period of time — we’re talking over a decade — and see whether it favorably affects sleep patterns, because we know that sleep patterns are also associated with the increased risk of cardiovascular death — either people that have sleep disturbance, people that have insomnia, people that sleep either too little (less than 6 hours) or too much (more than 9 hours).

These investigators had some data already. It was data from the European Community Respiratory Health Survey. They had obtained data points over a 10-year period. Two of those data points were: are you physically active — do you exercise at least 2 or more times a week, for an hour per week or more? Were you physically active when we started this study? Were you physically active 10 years later?

Oh, by the way, we also surveyed whether you had insomnia, whether you slept for more than 9 hours or less than 6, and did you have daytime somnolence?

Some individuals remain physically active over that 10 years; some started physically active and then subsequently became inactive, and some were inactive the entire time. Those individuals that maintained their physical activity over that course of 10 years, they were about 40% less likely to develop difficulties initiating sleep. They were more likely to have healthy sleep habits, less likely to sleep less than 6 or more than 9 hours, and there was no change in daytime sleepiness.

Those who started out active and became inactive — they didn’t receive any of those benefits. This is one of the few studies that’s looked over a long period of time to suggest that physical activity not only improves the other things I mentioned, but actually improves your sleep patterns as well.

Elizabeth: We know that this chronic sleep deprivation is really a major problem worldwide.

Rick: It is and it gets worse, by the way, the older you are. These individuals that they studied were an average age of 55 years ±7 and so we’re getting into that age group where sleep disturbances are pretty prominent.

Elizabeth: So get out there and exercise. If you are somebody who exercises, keep doing it.

Rick: Keep doing it and sleep well.

Elizabeth: On that note, that’s the look at this week’s medical headlines from Texas Tech. I’m Elizabeth Tracey.

Rick: And I’m Rick Lange. Y’all listen up and make healthy choices.

Please enable JavaScript to view the

comments powered by Disqus.