Diagnostic Performance of Various QTc Interval Formulas in a Large Family with Long QT Syndrome Type 3: Bazett's Formula Not So Bad After All …

Abstract

Background: Recently, we identified a novel mutation of SCN5A (1795insD) in a large family with LQTS₃. The aim of this study was to assess whether the various proposed corrections of the QT interval to heart rate help to improve the identification of carriers of the mutant gene.

Methods: The study group consisted of 101 adult family members: 57 carriers and 44 noncarriers (mean age 44.6 ± 14.6 and 40.3 ± 12.8 years, respectively). In all individuals a 12‐lead ECG, exercise ECG, and 24‐hour Holter ECG were obtained.

Results: Correction for heart rate significantly improved the diagnostic performance of the QT interval. Diagnostic performance of the Bazett formula was similar to that of the newer formulas (Fridericia, Hodges, Framingham, and a logarithmic formula). At a cut‐off value of 440 ms, the Bazett corrected QT interval was associated with a sensitivity and specificity of 90% and 91%, respectively. Using the 24‐hour Holter ECG, a prolonged QTc at heart rates less than 60 beats/min was almost pathognomonic for genetic mutation (sensitivity and specificity both 99%), whereas the QTc calculated at the lowest heart rate using Bazett's formula provided full discrimination.

Conclusion: In the present family, the resting ECG gave a good indication about the presence or absence of genetic mutation but a 24‐hour Holter recording was mandatory to ascertain the diagnosis. In the diagnosis of this form of LQTS₃, Bazett's formula was at least as good as other proposed corrections of the QT interval to heart rate.

QT prolongation has been recognized as a marker of increased risk for cardiac arrhythmias and sudden cardiac death. Already more than 35 years ago, the long‐QT syndrome (LQTS) was found to be a cause for sudden death in subjects with otherwise normal hearts. ¹ , ² More recently, several genetic variants of LQTS have been identified. We described a large family characterized by premature nocturnal sudden death due to a mutation (1795insD) in the cardiac Na⁺‐channel gene (SCN5A). ³ , ⁴ , ⁵ , ⁶ Carriers of this genetic mutation exhibit features of both LQTS₃ and Brugada syndrome. An important feature is the presence of excessive bradycardia‐dependent QT interval prolongation in conjunction with mild sinus bradycardia.

Although prolongation of the QT interval is obviously an essential characteristic of LQTS, identification of affected individuals may sometimes be difficult. In order to evaluate the QT interval under different conditions at varying heart rates, Bazett proposed in 1920 a correction formula in which the observed QT interval is divided by the square root of the RR interval. ⁷ However, Bazett's formula is said to lead to some overcorrection at higher heart rates and undercorrection at lower heart rates. Therefore, several other formulas have been presented, all aiming to better describe the QT‐heart rate (or RR interval) relationship. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ Nevertheless, Bazett's formula still is most often used for correction of the QT interval and, indeed, the existence of an “optimal” formula and even the necessity of any correction for heart rate have been questioned. ¹⁶ , ¹⁷

The aim of this study was to assess whether the various proposed corrections of the QT interval to heart rate help to improve identification of carriers of the mutant gene in the above‐mentioned family.

METHODS

The family came to our attention in 1958 when a 16‐year‐old boy was referred because of marked QT prolongation and biphasic T waves on an electrocardiogram (ECG). In view of a high familial incidence of unexpected nocturnal sudden death, clinical data were collected in subsequent years in as many family members as possible. Standard 12‐lead resting ECGs, 24‐hour ambulatory Holter ECG recordings, and exercise ECGs were obtained as part of routine investigations. At the time of the investigation, none of the subjects was using any cardioactive medication. Further details on the pedigree, clinical characteristics, electrophysiologic and genetic studies, follow‐up, and treatment have been given in earlier reports. ³ , ⁴ , ⁵ , ⁶ In this study, analysis was limited to adult individuals with definite genetic status, based either on DNA analysis ⁴ or pedigree analysis (obligate carriers).

Methods have been detailed previously. ⁴ , ⁶ Briefly, the standard 12‐lead resting ECG was recorded at a paper speed of 25 mm/s. The QT interval was measured from the beginning of the QRS complex to the end of the T wave where its terminal limb joined the baseline, taking the longest QT interval in any lead. From 1978 until 1988, 24‐hour ambulatory Holter ECG recordings were obtained using Avionics two‐channel 445 recorders and an Avionics 680 analyzer. Thereafter, Marquette three‐channel series 8500 recorders and a Marquette series 8000 XP analyzer were used. Recordings were made using modified leads V1, V5 and aVF. Measurements of the QT interval were made at the lowest heart rate and at fixed heart rates of 40, 50, 60, 70, 80, 90, and 100 beats/min, again taking the longest QT interval in any lead. Symptom‐limited exercise tests were performed using a bicycle ergometer. The protocol consisted of an initial work load of 50 Watts, and thereafter a stepwise 10‐Watt increase every 30 seconds. The 12‐lead ECG was continuously monitored using a Marquette Case 12 electrograph. A hard copy ECG was produced at regular intervals. Measurements of the QT interval were made at fixed heart rates of 110, 120, 130, 140, 150, and 160 beats/min and at the highest heart rate.

Corrections of the QT interval for heart rate were made in accordance with earlier reports. ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹³ , ¹⁴ , ¹⁵ Using Bazett's formula, ⁷ the corrected QT interval (QTc) is calculated as QT/(RR) ^1/2. QTc according to the Fridericia formula ⁸ is calculated as QT/(RR)^1/3, and QTc according to Hodges formula ⁹ is calculated as QT + 0.00175 × ((60/RR) − 60). Linear regression analysis using data of the Framingham Heart Study ¹⁰ yielded a linear adjustment QTc = QT + 0,154 × (1 − RR). Furthermore, we used an earlier described empirical formula, ¹¹ for which the parameters were optimized using the observations in the nongene carriers in our study. This resulted in the following formula: QTc = 0.402/(0.402 + (0.134 × Ln (RR))), which is further referred to as the “logarithmic” formula. In all the given formulas, values for the QT and RR interval are entered in seconds.

Data are presented as mean ± standard deviation, unless indicated otherwise. Differences between gene carriers and nongene carriers were assessed by the Student t‐test for unpaired samples. Fisher's Exact test was used for comparison of binary variables. Multivariate regression analysis was used to evaluate potential influences of gender and age. To evaluate the diagnostic performance of each formula in establishing the presence or absence of the genetic mutation in individual family members, Receiver‐Operator Curves (ROCurves) were constructed, in which sensitivity was plotted against 1‐specificity at varying cut‐off levels. For comparison of diagnostic performance, the Area Under the Curve (AUC) of each ROCurve was calculated—a larger area indicating a better diagnostic performance. All statistical analyses were carried out using SPSS for Windows, version 9.0.

RESULTS

The study group described in the current report comprised 101 individuals, 57 gene carriers (mean age 44.6 ± 14.6 years, 48% male) and 44 nongene carriers of the mutant gene (mean age 40.3 ± 12.8 year, 50% male). As expected, the QT and QTc intervals on the resting ECG were significantly longer for carriers as compared to noncarriers (Table 1). Similar results were found for QTc measurements on the exercise ECG (434 observations) and 24‐hour Holter ECG (625 observations). (Results not tabulated.) Again, the QT and QTc intervals were markedly longer in carriers, especially at lower heart rates. This difference gradually decreased with increasing heart rate. At a heart rate higher than 140 beats/min, the difference between carriers and noncarriers was no longer statistically significant. Using multivariate regression analysis, the possible influence of gender and age was evaluated. The calculated regression coefficients were small and statistically not significant for all measurements.

Table 1.

Patient Characteristics

	Carriers (n = 57)	Noncarriers (n = 44)	P‐value
				Age (years)	44.6 ± 14.6	40.3 ± 12.8	0.128
Male (%)	48	50	0.848
Resting ECG
Heart rate (bpm)	68 ± 15	72 ± 13	0.120
Uncorrected QT interval (seconds)	0.466 ± 0.095	0.377 ± 0.031	<0.001
QTc‐Bazett (seconds)	0.481 ± 0.044	0.409 ± 0.023	<0.001
QTc‐Fridericia (seconds)	0.475 ± 0.057	0.398 ± 0.019	<0.001
QTc‐Hodges (seconds)	0.479 ± 0.074	0.398 ± 0.019	<0.001
QTc‐Framingham (seconds)	0.474 ± 0.060	0.398 ± 0.019	<0.001
QTc‐Logarithmic (seconds)	0.477 ± 0.058	0.399 ± 0.019	<0.001
Holter ECG
Mean heart rate (bpm)	70 ± 8	77 ± 9	<0.001
Lowest heart rate (bpm)	41 ± 8	47 ± 8	<0.001
Highest heart rate (bmp)	124 ± 24	141 ± 16	<0.001

Values are mean ± SD.

The ROCurves are shown in Figure 1 (resting ECG), Figure 2 (exercise ECG), and Figure 3 (Holter ECG). The data on the accompanying AUC are given in Table 2. For the resting ECG, heart‐rate‐corrected QT intervals had a significantly larger AUC than the uncorrected QT interval (P < 0.001), indicating better diagnostic performance (Fig. 1, Table 2). However, the differences between the QTc intervals based on the five formulas were marginal and not statistically significant. For the exercise ECG, the AUC was lower for each QT and QTc measurement as compared to the resting ECG (Fig. 2, Table 2). Correction of the QT interval for heart rate according to the formulas of Bazett, Fridericia, and Hodges and the logarithmic formula resulted in better diagnostic performance as compared to the uncorrected QT interval. The difference in AUC between the uncorrected QT and QTc calculated according to the Framingham formula was not statistically significant. For the Holter ECG, the results were similar to those for the resting ECG (Fig. 3, Table 2). Again, heart‐rate‐corrected QT measurements had a significantly better diagnostic performance than the uncorrected QT interval (P < 0.001), whereas the differences in diagnostic performance between the five QTc measurements were negligible and not statistically significant.

Figure 1.

ROC curves for corrected QT interval—Resting ECG dashed line: uncorrected QT interval; solid line: reference line;  : QTc‐Bazett;  : QTc‐Fridericia;  : QTc‐Hodges;  : QTc‐Framingham;  : QTc‐Logarithmic.

Figure 2.

ROC curves for corrected QT interval—Exercise ECG dashed line: uncorrected QT interval; solid line: reference line;  : QTc‐Bazett;  : QTc‐Fridericia;  : QTc‐Hodges;  : QTc‐Framingham;  : QTc‐Logarithmic.

Figure 3.

ROC curves for corrected QT interval—Holter ECG dashed line: uncorrected QT interval; solid line: reference line;  : QTc‐Bazett;  : QTc‐Fridericia;  : QTc‐Hodges;  : QTc‐Framingham;  : QTc‐Logarithmic.

Table 2.

Corrected QT Interval in Relation to Genetic Status‐Area Under Curve of Receiver Operator Curve (95% CI)

Type of QT Correction	Resting ECG	Exercise ECG	Holter ECG
Uncorrected QT	0.876 (0.808–0.944)	0.612 (0.559–0.665)	0.809 (0.776–0.842)
QTc‐Bazett	0.953 (0.915–0.991)	0.720 (0.673–0.768)	0.963 (0.950–0.976)
QTc‐Fridericia	0.957 (0.918–0.996)	0.678 (0.628–0.727)	0.958 (0.943–0.972)
QTc‐Hodges	0.951 (0.909–0.993)	0.693 (0.643–0.743)	0.949 (0.933–0.965)
QTc‐Framingham	0.960 (0.922–0.998)	0.661 (0.610–0.711)	0.958 (0.944–0.973)
QTc‐Logarithmic	0.958 (0.920–0.996)	0.702 (0.653–0.750)	0.960 (0.946–0.975)

In the resting ECG, using cut‐off values of 400, 420, and 440 ms resulted in a sensitivity and specificity for establishing the presence or absence of the genetic mutation as listed in Table 3. It appears that the optimal cut‐off value, in which a high sensitivity is combined with a high specificity, was higher for the Bazett corrected QT interval than for the other QTc intervals. Using cut‐off values of 440 ms for the Bazett corrected QT interval and 420 ms for the other QTc intervals, a sensitivity and specificity of both approximately 90% could be achieved.

Table 3.

Corrected QT Interval in Relation to Genetic Status—Sensitivity and Specificity of Resting ECG

	400 ms		420 ms		440 ms
Cut‐off Value Type of QT correction	Sensitivity (%)	Specificity (%)	Sensitivity (%)	Specificity (%)	Sensitivity (%)	Specificity (%)
	Uncorrected QT	89	70	67	89	58	93
QTc‐Bazett	100	36	95	68	90	91
QTc‐Fridericia	97	55	93	91	74	98
QTc‐Hodges	97	57	91	91	65	98
QTc‐Framingham	98	59	93	91	68	98
QTc‐Logarithmic	98	55	93	91	75	98

Given the characteristic feature in this family of excessive bradycardia‐dependent prolongation of the QT interval, we additionally examined diagnostic performance at low heart rates using measurements from the Holter ECG. Using only observations at heart rates of less than 60 beats/min resulted in a sensitivity of 99% in combination with a specificity of 99% for the Bazett corrected QT intervals (cut‐off level 440 ms). Using the cut‐off level of 420 ms, comparable values for sensitivity and specificity were observed for the other formulas. Using the QTc at the lowest heart rate during the Holter recording led to a full discrimination between carriers and noncarriers of the mutant gene using the Bazett formula (i.e., AUC = 1, both sensitivity and specificity 100%). By contrast, such full discrimination could not be achieved using the other formulas (AUC ranging from 0.990 to 0.995).

DISCUSSION

To the best of our knowledge, this is the first study in which various formulas for correction of the QT interval to heart rate were evaluated in individuals with a genotypically established form of LQTS. A certain genotype of all members of this large family with features of both LQTS₃ and Brugada syndrome provided a unique opportunity to assess the diagnostic performance of each separate formula. The results indicate that correction of the QT interval for heart rate indeed greatly improved the ability to identify gene carriers in this family. In addition, the diagnostic performance of the “good old” Bazett formula was at least as good as that of the newer formulas that were introduced to improve heart rate correction of the QT interval. In fact, only using Bazett's formula in correcting the QT interval at the lowest heart rate during the 24‐hour Holter monitoring allowed full discrimination between carriers and noncarriers of the mutant gene.

Already in the report of Bazett ⁷ it was recognized that the QT interval at comparable heart rates is usually of longer duration in females compared to males. In contrast, we observed no differences in QT interval measurements between males and females. The absence of such a difference in our study is compatible with the finding of Lehmann et al., ¹⁸ who also observed no clear gender differences in their subgroup with LQTS₃.

In the present family, the QT interval on the resting ECG gave a good indication of the presence or absence of the genetic mutation but failed to ascertain the right diagnosis in all family members. Using only the resting ECG, roughly 10% of the carriers were incorrectly labeled as noncarrier (false‐negative) and also roughly 10% of the noncarriers were incorrectly labeled as carrier (false‐positive) of the mutant gene, irrespective of the formula used for heart rate correction of the QT interval. It is to be noted that patients in the present family were also characterized by changes in T‐wave morphology (like LQTS patients in general). ⁴ , ⁶ However, analysis of this phenomenon was beyond the scope of this study. Our data are in agreement with the previous finding by Vincent et al. ¹⁹ in other patient groups with LQTS that the value of the QT interval on the resting ECG is limited, in particular, regarding the possibility of false‐negatives. By contrast, analyses based on 24‐hour Holter ECG proved very accurate in this family. At heart rates of 60 beats/min or less, a prolonged QTc was thus almost pathognomonic for the genetic mutation, whereas the QTc interval calculated at the lowest heart rate using Bazett's formula actually provided full discrimination between carriers and noncarriers. Although in the exercise ECG differences in QT and QTc intervals between carriers and noncarriers remained statistically significant up to 140 beats/min, the diagnostic performance of the exercise ECG was limited in comparison with the resting ECG and Holter ECG. Still, the diagnostic performance of the Bazett formula was again at least as good as that of the other formulas.

The observation that the optimal cut‐off value for QTc using Bazett's formula is higher than for the other QTc formulas may be explained by earlier observations that Bazett's formula leads to some overcorrection at higher heart rates and undercorrection at lower heart rates. Given the fact that for most observations in this study heart rate was higher than 60 bpm, on average some overcorrection is to be expected. As discussed above, this suboptimal correction of QT interval to heart rate apparently does not adversely influence diagnostic performance. Although it is clear that some form of correction of the QT interval to heart rate is important, one should view the magnitude of change induced by the various formulas in comparison with other well‐known sources of variation in the QT interval such as autonomic nervous system activity, posture, diurnal variations etc. ²⁰ , ²¹ , ²²

CONCLUSION

In this particular family, the QT interval on the resting ECG gave a good clue about the presence or absence of genetic mutation, but a 24‐hour Holter recording provided almost full discrimination. In addition, Bazett's formula was at least as good as other the formulas to correct the QT interval for heart rate. Whether these observations also apply to other forms of congenital LQTS (particularly those with relative QT prolongation at higher heart rates) or in acquired LQTS remains to be established. For the time being, Bazett's formula does not seem so bad after all.