Robert R. Fenichel


Frequently Questioned Answers (2)

5. Electrocardiograms

5.1. What is measured by an electrocardiogram?

When cells depolarize, they become positively charged with respect to their non-depolarized neighbors.  The positive charge can be measured as a positive voltage between points on the surface close to the affected cells and points on the surface farther away.

Body-surface measurements can detect electrical activities of nerves and skeletal muscles, but when a subject is at rest and leads are appropriately placed, the body's dominant electrical activity is that of the heart.  The electrocardiogram (ECG, but sometimes appearing as EKG from the German) records the change over time in surface voltages arising in the heart.

5.2. What is the electrocardiographic appearance of a typical cardiac beat?

This drawing is adapted from Netter FH, The CIBA Collection of Medical Illustrations: Heart (Summit, NJ: CIBA Pharmaceutical Company, 1969), page 50:

This appearance is of course specific to one set of leads.  For example, if the leads were exactly reversed, the tracing would be exactly upside down.  The sequence of excursions would remain the same: P wave, PR segment, QRS complex, ST segment, T wave, and U wave.  The time between one beat and the next is often referred to as the RR interval.

5.3. To what action potentials do the various portions of the ECG complex correspond?

When parts of the atria are depolarized and parts are not, the net voltage is detected as the P wave.  At the same time and during the ensuing PR segment, depolarization is progressing through the AV node and conduction system, but these tissues have insufficient mass for their activity to be detected.  Progression of depolarization through the ventricles is seen as the QRS complex.  Once depolarization of the ventricles is complete, there is no net voltage across the heart, so the tracing returns to its baseline during the ST segment.  As repolarization of the ventricles proceeds, however, the heterogeneities within the ventricular wall become important, and repolarization is for a while more complete in some cells (those of the endocardium and epicardium) than in others (those of the mid-myocardium).  The resulting voltage gradients are manifest in the electrocardiogram as the T wave.

The origin of the U wave was obscure for many years.  Its normal timing corresponds to that of repolarization of the Purkinje fibers, but the Purkinje fibers are insufficiently massive to account for large late-appearing waves that are often described as U waves.  These large waves are more likely to be late components of a prolonged T wave, generated as described just above.

5.4. How are the durations of the various intervals affected by heart rate?

The PR interval (normally 120-200 ms) and QRS interval (normally 80-120 ms) are not much affected by changes in heart rate.  The QT interval shortens as the heart rate is increased, and this phenomenon is central to the whole next section.

6. The QT Interval

6.01. How do we decide whether a biometric datum is “normal”?

Some physiological measurements have a single normal range, independent of the subject's age, sex, or other variables.  For example, everyone's serum sodium concentration is normally in the range 135-145 mEq/L.

Other measurements are more complicated, because they vary with one or more important cofactors.  The question “What is the normal range of adult weight?” is not meaningless, but the answer (something like “100-200 pounds”) is almost uselessly imprecise.

We sometimes decide whether a measured body weight is normal by referring to tables that show a weight range for each of many combinations of sex and height.  Rather than deal with the bulky tables, physicians sometimes use simple rules of thumb, like “a normal adult woman 5 feet N inches tall weighs about 110+4N pounds.”  Finally, it is sometimes convenient to run rules in reverse, normalizing the measured data by deriving a metric sometimes called an index.  In the case of weight, the conventional calculation is that which yields the body-mass index, BMI = weight / height2. The claim of the BMI is that once it has been computed, the normality or abnormality of the subject's weight can be assessed with no further reference to his or her height.

All of these schemes must be evaluated, at least at first, in the same way. A useful scheme should be unbiased, and to be most useful it should be reasonably precise.

A scheme is unbiased if it actually describes the population for whom it is a proposed standard.  To say that the normal range for adult human weight is 100-200 pounds may not be terribly useful, but it is (approximately) correct.  A suggested range of 300-350 pounds would be more precise, but it would be biased.  A given range, table, rule, or normalization might be unbiased for one population, but biased for another.  For example, the rule of thumb given earlier functions well for short women, but the only tall women it accepts as “normal” are unnaturally thin.

Although a normal range can be meaningfully defined for any scheme that is not seriously biased, the range may need to be uselessly imprecise if the scheme does not take account of pertinent cofactors.  This was demonstrated in the example of the single normal range for adult weight.

6.02. What is meant by QT “correction”?

A QT “correction” formula is a formula that combines the measured QT duration and the measured heart rate into a computed index, similar in concept to the BMI.  The term “correction” is a misnomer, inasmuch as the measured QT interval is not incorrect (as a weight might be incorrect if it were measured while the subject were wearing heavy clothing).  Nevertheless, the term “correction” is deeply entrenched, and the output of the calculation is usually called the “corrected QT interval,” denoted by QTc.

6.03. What formula is used for QT correction?

Several dozen different formulas have been proposed over the years.  The most frequent style of these formulas has been

QTc = QT / RRa

where a is a formula-specific parameter, and of these formulas the one most frequently seen is the formula attributed to Bazett, in which a = 0.5, and for which the published upper limit of normal is 440 ms.  In what follows, the results of correction using the Bazett formula will be denoted by QTcB.

6.04. Is the Bazett correction unbiased?

No.  Bazett studied only a few electrocardiograms in each of a small number of subjects, and he then chose a = 0.5 by subjectively observing the fit to his data.  In more recent studies, using dozens of electrocardiograms from each of hundreds of subjects, the best value of a (by formal statistical methods not known to Bazett) has varied from population to population.  Values as high as 0.5 are almost never seen; the most frequent values are slightly below 0.3.

Using an excessive value for a causes one to expect that with increasing heart rate, the QT interval will decrease more rapidly than it actually does.  As a result, normal reductions of QT duration with heart rate will be perceived as insufficient, suggesting that some concomitant process is prolonging the QT interval.  With decreasing heart rate, an excessive value of a will lead to acceptance as normal of QT durations that are actually abnormally long.  For example, Browne et al. (Am J Cardiol 50(5):1099-1103 (1982)) studied volunteers who received atropine.  Atropine normally raises the heart rate, but in one phase of the trial the volunteers’ hearts were paced to prevent any heart-rate change.  As shown in the figure, the measured QT interval was shortened from baseline by 

24 ms; this is the true effect of atropine on the QT interval at that heart rate.  In another arm of the study, the same volunteers received atropine in the absence of pacing.  The measured QT interval was now decreased by 49 ms, combining the true effect (24 ms) with an extra 25 ms induced by the rate change.  An adequate “correction” formula would have adjusted for the rate and come out with a decrease close to the gold-standard 24 ms (probably somewhat larger than 24 ms, since the correction is multiplicative, and the paced rate was probably faster than 60 bpm).  The Bazett correction (a = 0.5) computed that these subjects had suffered a 43-ms increase in their true QT intervals.  This was nonsense. 

6.05. When the heart rate is close to 60, so that QTcB is close to QT, need I be concerned about the failings of the Bazett formula?

Probably, yes. The accepted upper limit of normal for QTcB (and for QT when HR is around 60) was probably not obtained by averaging the QT intervals of subjects whose heart rates were all around 60 bpm. Instead, the normal range was probably obtained by averaging the QTcB results of subjects at various heart rates, most of them higher than 60. Because these QTcB values were mostly too high, the accepted upper limit of normal is probably too high.

6.06. Hasn't the Bazett formula been validated by years of use?

Yes, Ptolemaic astronomy has been....  What?  Sorry.  Yes, homeopathy has been....  Oh, sorry again.  I get distracted.  Please be so kind as to repeat your stupid question.

6.07. Would it be better to use a similar formula with a around 0.3?

For most populations, yes, but studies have found population-specific values of a ranging from less than 0.2 to greater than 0.6.  It's better to pool your baseline data to find the a that best fits your specific population.

Even if you do that (or if you choose an arbitrary value and it turns out to be a good fit), you will probably find that although bias has been minimized, precision remains mediocre.  That is, individual variation in the QT/RR relationship will probably be so large that a population-specific normal range will need to resemble the height-independent normal range for weight discussed much earlier.  You may get slightly better results by considering formulas different in style (say, using Taylor series), but the intersubject variation will probably remain a problem.

6.08. Would it be better yet to obtain sufficient baseline data to fit a separate formula for each subject?

Yes.  With this technique, it is possible to detect abnormal QT intervals, regardless of heart rate, with high precision.  Techniques for doing this are described in the paper cited above.

6.09. When the heart rate changes, does the QT interval change right away?

No.  After a change in heart rate, the QT interval takes a minute or two to reach its new steady state.  Thus, for example, a QT interval measured just after an increase in heart rate may falsely appear to be abnormally long, because it is still shortening from its (normal) duration before the change in heart rate.  Again, see the paper cited above for further discussion.

6.10. What is meant by "QT dispersion"?

QT dispersion was at one time thought to be the electrocardiographic manifestation of important variation of AP duration in ventricular muscle.  Depending on their degree of mathematical sophistication, different authors identified QT dispersion with different measures (standard deviation, maximum-minimum difference, etc.) of the range of QT duration as seen in the various electrocardiographic leads.  This notion of QT dispersion was popular a few years ago, but it lost momentum when it could not be shown to be of independent prognostic significance.  These measures' insensitivity should (in hindsight) have been expected, inasmuch as each of the proposed measures of QT dispersion was sensitive to differences in milieu between regions of ventricular tissue that were physically far separated from each other, and therefore unlikely to join in hosting reentrant arrhythmias.

Intraventricular variation in AP duration is important, but the important kind of variation is radial rather than circumferential.  That is, the important variation in AP duration is the variation seen within a localized slab of ventricular wall, among the epicardium, the mid-myocardium, and the endocardium.  This sort of variation is present in normal hearts (in which it generates the normal T wave) but it is also the sort of variation that is magnified by drugs that induce torsade, and not by drugs that even if they prolong the QT interval do not induce torsade.  Unlike the circumferential variation, unfortunately, the radial ("transmural") variation in AP duration is not directly evident on the surface electrocardiogram.

            previous page                    back to questions                              next page

Page revised: 11/29/2010 21:48