In the musical “My Fair Lady,” Professor Henry Higgins rhetorically asks: “Why Can’t a Woman Be More Like a Man?” A somewhat similar question haunts scientists interested in the assessment of cardiac function during atrial fibrillation: why cannot we make the left ventricular measurements in atrial fibrillation be less variable, as they are during sinus rhythm? And there are several reasons why is it difficult to say “yes” to this question.

Atrial fibrillation can be viewed as a series of beats that are both extrasystolic and post extrasystolic; that is, in atrial fibrillation, ventricular contractions show both the effect of extrasystolic restitution and post-extrasystolic potentiation. The result of this is quasi random variability of left ventricular (LV) function (as quantified by systolic or diastolic function parameters such as ejection fraction, stroke work, or the time constant of isovolumic pressure decay). Mathematical modeling has shown that much of this variability can be accounted for by using the ratio of the preceding and pre-preceding intervals (RR1/RR2): the larger the RR1/RR2 ratio, the better contraction (or relaxation) is ( Figure 1 ). In other words, an RR1/RR2 of 2.0 would lead to a more powerful contraction that RR1/RR2 of 0. 5. While the relationship between RR1/RR2 ratio and ventricular response is curvilinear, and ventricular response still also depends also on the average heart rate, and ventricular response still also depends also on the average heart rate and LV preload (through the Frank Starling mechanism and the force-frequency relationship), it can be efficiently approximated using a linear relationship. More important to the topic here, mean ventricular function during atrial fibrillation can be quantified by measuring ventricular function of a single beat whose RR1/RR2 equals one. In other words, when preceding and pre-preceding intervals are equal, thus making atrial fibrillation more comparable to a sinus rhythm.

Relationship between the ratio of preceding and pre-preceding beats (RR1/RR2; x-axis) and theoretical ventricular systolic performance calculated using equation of Suzuki et al (y axis). Data are derived from a real-life series of 500 consecutive RR intervals obtained during atrial fibrillation. While the data are non-normally distributed and the relationship between ventricular performance and RR1/RR2 is curvilinear ( broken pink trendline ), this relationship in the region of R1/RR2 values between 0.75 and 1.5 can be approximated by a straight line ( full black trendline ), with the exactly one-half of the data points being above or below the value of RR1/RR2 = 1.

In the current issue of the Journal , two manuscripts assess the feasibility of substituting a measurement of ventricular function during a single beat with RR1/RR2 ratio being equal (or close) to one, instead of measuring and then averaging multiple consecutive beats. Lee et al assessed 98 patients with rate-controlled atrial fibrillation. They obtained and calculated global longitudinal left ventricular strain from 15 consecutive cardiac cycles, and showed that essentially the same strain value could be obtained by measuring just one cardiac cycle if the preceding and pre-preceding RR intervals did not differ by more than 60 ms. This is a great help for the routine practice of echocardiography in atrial fibrillation as sonographers often face the dilemma of either not performing measurements, or measuring multiple cycles without being certain how representative these cycles are of overall state of cardiac function. It also shows clinical support for previous experimental work. Lee et al also showed that patients with higher CHADS2 scores had lower longitudinal strain. Such findings are expected, as patients with higher CHADS2 scores have borderline lower ejection fraction, and, by definition, have higher frequency of heart failure, diabetes and hypertension, all factors that affect longitudinal strain.

A slightly different approach to the problem of measuring systolic and diastolic left ventricular strain was described by Kusunose et al. The authors simultaneously acquired color tissue Doppler echocardiography and micromanometer-recorded left ventricular pressure data over 10 to 20 cardiac cycles in a series of 25 patients with atrial fibrillation, with the aim of exploring the associations between hemodynamic and echocardiographic variables of strain and systolic and diastolic strain rates. As a next step, they showed that strain and strain rates were influenced by RR1/RR2 ratio, and finally they again validated the concept that a ventricular response at an RR1/RR2 ratio of 1 is representative of the average ventricular response over a series of successive complexes during a 10 second period of recording. This is an interesting study for several reasons. Correlations between invasive micromanometer pressure data and echocardiography parameters are rare. Micromanometer catheters are expensive, and data can be acquired only prospectively; however, they provide hemodynamic data that is otherwise not available, such as peak positive and negative derivative of LV pressure and time constant of isovolumic pressure decay (tau). Peak positive and negative derivatives of LV pressure are preload dependent measures of LV contractility and relaxation, respectively, while tau is the gold standard measure of LV relaxation. Kusunose et al also showed a within- and between-patient association between peak positive derivative of LV pressure and systolic strain, between peak negative derivative of LV pressure and diastolic strain rate, and between tau and diastolic strain rate. It is important to note that these findings show associations – both invasive and noninvasive parameters are influenced by RR1/RR2 ratio, however the current body of knowledge suggests that these relationships are mechanistically determined.

There are some important differences between these studies. Lee et al used a speckle tracking method to determine strain, which is both more accurate and more tedious than the color tissue Doppler method. On the other hand, given the lower sampling rate, they do not report on strain rates. Kusunose et al used a color Doppler tissue method to calculate strain, which has better sampling resolution, is easy to apply, but very operator sensitive. Not surprisingly, global longitudinal left ventricular strain was higher in the study of Kusunose et al (-17±2% vs.12±3%; P <.00001) despite patients having similar ejection fraction.

As is common to all scientific investigations, the studies presented here also carry some limitations. Most of the patients had well controlled heart rates, with average of 76±17 bpm in the study of Lee et al, and 74±15 bpm in the study of Kusunose et al. Within-patient RR variability is not reported, and we can only guess how irregular was atrial fibrillation (AF) in patients studied. While in the clinical setting most of the patients with AF do have a controlled heart rate, assessing systolic function during AF with rapid ventricular rate may be especially relevant. On the other hand, AF with rapid ventricular rate in itself produces hemodynamic compromise even when contractility is normal, so from the point of view of treatment the obvious thing to do is to slow the heart rate. Also, depressed contractility in the presence of rapid ventricular rate may also be transient-either due to tachycardia-induced or catecholamine related left ventricular dysfunction, and it can change dramatically with rate (or rhythm ) control. The two studies presented here also emphasize the need for standardization of new methods, such as strain. Clearly, what would be considered as abnormal by color Doppler derived strain, would be classified as normal by speckle tracking imaging. These issues are becoming even more pressing with multiple vendors and multiple imaging modes continuously providing the echocardiographic community with new and untested options.

What is the physiology behind these observations? Why is ventricular systolic and diastolic behavior on a beat following two matched R-R intervals representative of “average” behavior over a longer time frame? Figure 1 may explain why. It shows a theoretical ventricular systolic performance derived from a real-life series of 500 consecutive AF RR intervals using the equation of Suzuki et al. While data are non-normally distributed and the relationship between ventricular performance and RR1/RR2 is curvilinear, this relationship in the region of R1/RR2 between 0.75 and 1.5 (with highest density of data-points) can be approximated by a straight line, with exactly one half of the data points being above or below the value of RR1/RR2 = 1. Thus, one can substitute ventricular performance of a beat for which RR1/RR2 approaches one for a mean (or more precisely a median) measure of ventricular performance.

In the end, if selecting beats with an RR1/RR2 ratio of one is so practical, why hasn’t it been adopted more widely in the assessment of patient with AF? There are multiple reasons for that. Most software approaches do not support measurement of RR interval while in a two-dimensional image mode, and thus identification of a beat with RR1/RR2 of 1 is difficult. Also, the impact of average heart rate re-emerges: measuring studies with substantially different cycle lengths is not physiologically meaningful as heart rate impacts cardiac function, whether the patient is in AF or in sinus rhythm. So detecting poor contractility or relaxation in a patient with AF and heart rate of 130 bpm does not necessarily mean that this same patient will have poor contractility once his heart rate is below 90 bpm. With all these caveats, with increasing prevalence of AF cardiologists ought to be aware that the shortest way to accurate and stable assessment of cardiac function in this setting is to measure the cardiac beat just after two cardiac cycles of similar length.