SEGMENT 2: Dr. Freiberg discusses some of the most intriguing findings from his Big Data work.

Dr. Gary H. Gibbons, Director, National Heart, Lung, and Blood Institute:  We have talked about the methodologies and platforms. Give us a little bit of an update and state of play of what you are most excited about in some of the observations you have made, leveraging these systems and some of these big data opportunities you have been roving through. What are you finding out that is exciting?

Dr. Matthew Freiberg, University of Pittsburgh:  One of the things that we have been able to do in the VA, because we have these large numbers, is to look at HIV-infected people first with hard events. Unlike using subclinical measures of disease, whether coronary calcium and subclinical atherosclerotic, and don’t misunderstand me, those are really important measures. But it would be nice to also know if, do these people have heart attacks or do these people have increased risk of stroke, or do these people have increased risk of heart failure? 

With the data that we have put together with the help from the VA, and there’s been a lot, we have hard events, real AMIs [acute myocardial infarctions], and not just ICD-9 codes, but the AMIs, their enzymes and EKGs, just like you might adjudicate in Framingham. With our large sample size, not only have we been able to show that we think HIV carries an excess risk of 50 percent, but because the sample size is so large, we can look at subsets of people in a way that you might not be able to otherwise. 

And because we have all their clinical data, for example, we can look at how HIV viral load changes over seven years and how that impacts care. For example, in our data set, we showed that even if you achieved a viral load of less than 500 consistently and sustained over a course of years and years, these people still had an excess risk of almost 39 percent for having heart attacks. So even if you are taking your meds and you are driving that viral load down, there is still that excess risk. 

Well, why is that? Is that a function of the virus is replicating somewhere we cannot see and creating an immune response? Is it the damage the virus has already done, and even though you’re getting the virus under control, we still need to think about clever ways to minimize the impact? Is it some of the comorbid disease interacting with HIV? That’s where we’re trying to understand what’s happening there. 

We have also been able, because the sample size is so large, as we mentioned earlier, no it’s not smoking, as an example. We have been able to restrict our sample size down to 15,000 people, but it’s still enough to say, even among never smokers, we’re still seeing this risk. Then using the tool that the NHLBI helped us build, now we know that HIV infection is associated with not only, not just heart failure, but both preserved and reduced ejection heart failure. That’s kind of important. Because if you have both of those, it’s not necessarily HIV increasing atherosclerosis that’s driving heart failure. It’s entirely possible that HIV may be doing something directly to the myocardium that is translating into this heart failure risk. If that were true, then we need to do more than just reduce atherosclerotic risk factors. We may need to be thinking in a slightly different way or maybe even a bigger way. 

We’re seeing similar findings with stroke, ischemic stroke, too, suggesting that both CHD [coronary heart disease] and stroke, it’s a similar mechanism. For us, we feel like having these big data sets now are telling us that we think these associations really translate into increased risk of hard events. We’re understanding that we don’t think it is all the comorbid disease. We really think it is the virus and the virus or the treatment. The fact that the VA has given us all their data, we have all their medication data, every prescription refill, every laboratory value that they’ve ever had. The ability to have access to those data, and in a de-identified fashion, and we should touch on that in a second, allows you to look at these associations. 

The other thing that’s important, as I think about it, but I want to bring it up, is that some people may be really concerned about well, this doctor can see all my data. This makes me feel uncomfortable as a patient in this health care system. I don’t. How these data sets get built with the big data to look at these is that people within the VA central office, we give them the variables we want and they merge all these data together and then provide me the data sets to use. I don’t look at someone’s name or an individual chart, per se, when I see the data. We have other people do that so it protects the integrity of the process. That’s really important to understand.