3.1 Review of HRV in Epidemiological Studies

• The title of this talk is a meta perspective on heart rate variability, antiquated or indispensable. As I mentioned in my first two talks in this series, this is the way I interpret how HRV has evolved over the past 50 years and I divided the five stages, understanding HRV HRV as a marker of autonomic nervous system HRV is a marker of body weight function HRV and the construct of Mind Body interaction, and HRV itself is a desirable target. Why these categories have been divided accordingly will become clear when this top series is over. But it reflects the evolution of our scientific approach to HRV which has been strongly influenced by the prevailing scientific paradigm of the day.
• So when we talk about the first stage of understanding HIV, this occurred in the 1970s and 1980s, which I believe was the strong age for physiology in general. As a consequence, this is where we begin to understand what biological processes are contributing to heart rate fluctuations, which includes respiratory sinus arrhythmia, barre, flax and other physiological processes. And this also was a time when you identify the signaling pathway involved in heart part rhythmicity and the role that the autonomic nervous system plays in modulating it. This is the time when heart rate variability, time domain and frequency domain measures are established and formalized. And this is a time where we also recognize that higher frequency HRV is associated with the parasympathetic branch predominantly, and the low frequency is a combination of both sympathetic and parasympathetic nervous systems.
• The 1980s and 1990s was the age for major advances in cardiovascular medicine. This was a time when the autonomic nervous system was recognized as playing a pivotal role in the occurrence of deadly arrhythmias or sudden cardiac death, particularly post MI. And heart rate variability as a proxy for autonomic nervous system activity was used to risk stratify post mi patients. And in fact, HRV showed great promise as individuals with lower HRV had markedly higher mortality.
• But with the advent of rapid revascularisation and beta blockers, HRV became relatively obsolete. Plus, it was messy, it is confounded by multiple factors as I had mentioned in part two of this series. So within this context, we are now in the 1990s and 2000s, where epidemiology nonlinear scientists have come to the fore. And I believe that both these disciplines help reframe HRV as a marker of body weight function. And hopefully by the end of the talk, you get an understanding why I say that.
• The goals of this talk are to review HRV epidemiological studies, understand basic concepts underlying nonlinear dynamics, appraise evidence for fractal base HRV measures, and summarize other nonlinear HRV measures. We'll start off with reviewing the HRV epidemiological studies.
• I think we all know what an OT large observational cohort study is. An example of this is the Framingham Heart Study, which enrolled its first patient in 1948. Typically, a large number of people are recruited from a community and observed over time, and to assess how a risk factor or exposure affects a specific outcome of interest. In our case, we want to see how Heart Rate Variability measures like SDN N relates to development of CV event, cardiovascular event or death.
• So the question I put to you is, when do you think these cohort studies get published? If the golden age of cardio of medicine, Cardiovascular Medicine began in 1980s, when do you expect to see the epidemiological studies to come out?
• To answer that question, I put here the list of major milestones for the Framingham Heart studies. And you could see that some of the major discoveries that we know now as pretty obvious, occurred in the 1960s, such as cigarette smoking is associated with heart disease or high cholesterol, high blood pressure associated with heart disease. You see, some of the results sort of trickle in, in the 1960s, and 1970s, and 1980s. But it was really in the 1990s that it really picked up where major milestones were discovered almost on a yearly basis. And if you recall, clinical cardiology made advances in 1980s. So there's usually a 10 to 20 year lag before advances in the clinical world makes its way into epidemiology. And a lot of it has to do with the time it takes to recruit those participants to observe the cohorts and also some practical implementation of some of the new technologies. In the case of heart rate variability. It was the Holter technology.
• So this was one of the first studies that came from Framingham. This was published in 1996 by Tsuji. The title was impact of reduced heart rate variability on risk for heart cardiac events. They recruited 2500 individuals, the mean age was 53 years of age, the participants were free of cardiac history, and they obtained two hour ambulatory EKGs. During the clinical visits and took a look at three time domain and frequency domain measures. The primary outcome was cardiac event, which was defined as angina unstable angina myocardial infarction, CHF, or sudden death or non sudden cardiac death. And there was a total 58 cardiac events. And the statistical analysis involved Proportional Hazards regression analysis, it was adjusted for age, sex and other clinical measures which were considered important for cardiac mortality or risk.
• This is the table that came from that study. And there's two things that I want you to notice. First, the heart rate variability is presented in logarithmic form. And the reason they did that was because the heart rate variability data itself was skewed logarithm enabled it to be normalized and and to be able to be applied in the statistical analysis. And the other thing is, there's this dash 100 for these two variables. And what they did was they divided the two hour recording of the ECGs into 102nd segments. And so the very low frequencies and the total power are of those 102nd segments, and not necessarily of the whole two hours. And we see here that a few of these variables are significant, including SDN in very low frequency, low frequency, high frequency and total power. However, the one that's associated with the greatest significance is the two are SDN.
• So what physiological process does the two hour correlate with, and this is a slide that I included in part one of my series, and it answers the question what physiological process discussed again, and most correlate with and it really depends on the duration of the data. And the reason for that is that heart rate variability is distributed largely in the lower frequencies when it comes to power spectrum. So the longer the data that you collect, the more it's going to predominate by by the lower frequencies. So two minutes associated with high frequency five minutes with low frequency and high frequencies, one hour is very low frequency and 24 hours is ultra low frequencies. So the two hours from the Framingham study is more associated with very low frequencies.
• The reason the very low frequencies is relevant is because at least when I had put this slide up, in part two of the studies, we saw that very low frequencies was highly correlated with mortality after EMI. And this core graph clearly shows that the variable frequency was much more significant compared to the other higher frequencies. And it also showed in this table here, where the ultra low frequencies was associated with all cause death, whereas our cardiac and death was associated with very low frequencies.
• The other thing that I wanted to mention from this study was that this was a two hour ambulatory ECG acquisition during clinic visits. This is in significant contrast from what I had mentioned in the part two of this series. In part two, there was a strong emphasis on avoiding factors that influence HRV. I had mentioned that there were many cardiac confounding factors, such as you know, food intake time of the day, whether you had even drank water or whether your bladder was full. And as a result of that, there was a lot of instructions given to patients and still are regarding what they can or cannot do before they undergo autonomic test that included no food for three hours beforehand and alcohol for 12 hours before.
• This very rigid, regimented protocol is in stark contrast to the approach that was taken in Framingham study because the individuals, the participants, when they came to the clinic, could actually do whatever they wanted. They could go walk around and they participant in various studies, and so there was not no control for these potential confounders. Also, the time of day wasn't exactly recorded. So this could have been done at any time of the day. This dichotomy in the way of ECGs obtained really highlights is the differences in how HRV is interpreted. A good example, this is something that I encountered myself in collaboration with Harvard OSHA Research Center. We were interested in seeing whether Tai Chi masters had higher heart rate variability compared to healthy controls. And they indeed had the higher heart rate variability, particularly our MSSD.
• It was found out the reason probably for the higher heart rate variability was due to the fact that they took slower and deeper breaths. But this led to differences in the interpretation of the higher heart rate variability. One group said, well, the greater Heart Rate Variability does not indicate better health because it is confounded by bigger, slow breaths. Another group, however, said, greater Heart Rate Variability indicates better health because he or she takes bigger slow breaths. So you can see the differences in the way heart rate variability is interpreted. One group says a higher heart rate for abilities confounded by different practices. Whereas another group says the higher heart rate variability is actually reflection of better habits. And this extends not only to breathing but also to physical activity and good sleep as well. So you can see that this differences started to manifest into the 1990s and 2000s, because there's a different type of data that was obtained or used for heart rate variability, one that was very controlled for cardiology, and autonomic neurology, another one that was pretty much free running and real life conditions.
• The other thing I would like to highlight from this tissue study in 1996, was the participants. There was in the end 2501 participants. But if you take a look at the table, one of the study, there was up to 3420, who actually had ambulatory ECG data. But nearly a fourth were excluded many because they had ectopic or non sinus rhythm beats. And the reason I wanted to point this out was because expect to exclude up to a fourth of the patients that you enroll in your studies, if you're going to evaluate heart rate variability, and for many of the subsequent studies that will be cited in this talk, have excluded atrial fibrillation, and it's one of the weaknesses of heart rate variability is that you do have a large segment of the population excluded for these reasons.
• This is a meta analysis done by Hillenbrand in 2013, which summarizes the studies that looked at heart rate variability in a general population. And we can see here that it was done in numerous locations with a focus largely on SDN N, with multiple endpoints, but mostly about cardiovascular death. And there's also others with myocardial infarctions, other quality outcomes. Many of these were large studies mean follow up range from 3.5 to 13 years and a good range in age. But one thing I wanted to point out was the duration of measurements. And the reason why this was important is that these measurements are quite short. And when you have different durations, and as mentioned in SDN slide, is that the shorter duration indicates higher frequencies, whereas the larger ones are associated with lower frequencies. And so the summary here does not truly indicate the same physiology processes going on. The shorter segment ECGs are also quite sensitive to a number of confounding factors. As I had mentioned in my first part of the series is that high frequencies and low frequencies has a circadian rhythm, for instance, or it can be affected by stress or other factors. So taking a snapshot of 10 seconds to 30 seconds or a few minutes, in a single day will just take only a snapshot when so it is unclear how effective these measures will be compared to much longer term recordings.
• This is the forest plot that shows the effect of low heart rate variability compared to Heart Rate Variability when we're regards to fatal and non fatal cardiovascular disease. And if we categorize according to the duration of the measurements we see, for these studies, which were acquired data for less than three minutes, you see that the relationship in terms of premature mortality is not as strong. Whereas for those studies that acquired much longer data for two hours, or 24 hours for both these studies, the relationship is much more consistent. And the NACA collio may not be as consistent because it was a much smaller study. Already variability also has associations or important prognostic informations for post MI and congestive heart failure as well not only in the general population. And so the summary of the epidemiological HRV studies is that HRV has significant prognostic implications even in free running conditions.
• Lower heart rate, variability, frequencies are very important. And that's a factor that I think we've ignored over the past few decades. Very low frequencies and ultra low frequencies in particular have not had received much attention as of late and harbor variability is applicable to the general population and not limited to specific cardiovascular conditions. And nearly all studies dysrhythmias such as a fib, frequent ectopic beats were excluded. And various conditions such as use of beta blockers post cabbage may limit HRV usefulness, and I didn't really go into the details, because we didn't have time to go over this in a one hour talk. But post cabbage particularly can significantly diminish Heart Rate Variability even up to months afterwards.
• Nearly all the studies acquired only a snapshot of heart rate variability, for instance, in a single day or even part of a single day. So it is surprising to me that we even found a significant result in these studies. A two minute snapshot in one day four years can really it's hard to imagine can really impart any information about mortality.
• And monitoring dynamic changes in heart rate variability are potentially helpful but understudied. And this is an example where the dynamic changes can really be important. This is a study done by Adamson in 2004 and 288, CHF congestive heart failure patients who had very advanced congestive heart failure. They had an implanted cardiac resynchronization device and this device is able to track the atrial beats and to see if there was an actual normal sinus rhythm.
• What they did is they evaluated a certain heart rate variability and measure called SDA M is the standard deviation of five minute median atrial interval is this type of very low frequency measure. And they visit this across the days and they've done this up to two to three months and even longer. And they've noticed that when it came to the hospital admission, the heart rate variability measure decreased steadily right before the hospitalizations indicating a route towards decompensation. They also saw a rise and the night heartbeats and also a decrease in the patient's physical activity. He created a detection algorithm for this decomposition and CHF. Essentially, they took a mean, and then found the periods where they had consecutive decreased below the mean. And you had accumulated difference and then decided what threshold would be to activate this D compensation algorithm. And this D compensation algorithm would did a great job in detecting those individuals who ultimately got hospitalized. It did a better job than physical activity, and even night heart rate.
• And finally, modulation of higher heart rate frequencies has been effectively ignored. This was a study that I cited in the part one of my series. This was a comparison of post mi patients those with high heart rate variability and low or heart rate variability. The individuals with high heart rate variability had higher heart rate variability because they had significant circadian modulation of the high frequencies throughout the course of the day. Whereas the low heart rate variability individuals did not. And this modulation of the high frequencies and low frequencies of heart rate variability is largely ignored now, and it's something that I think is worth evaluating again, and importantly, the modulation of high frequency heart rate variability is not captured in the very low frequencies also low frequency amplitudes

‍