Absolute QT interval was measured by hand in leads IIand V5. When II or V5 measurements were impossible, the interval was taken from V4 or V6 or any lead in which it was more readily defined. Automatic measurement was not considered for 2 reasons: a) it was not available in all electrocardiographs used in the study and b) electrocardiographs equipped with automatic measurement options used different algorithms. The Bazett formula (QTc = QT/√RR) was used to calculate the corrected QT (QTc). Although this formula has limitations (tendency to overestimate QT above 100 beats), it is the most used commonly formula and, therefore, the results can be compared with those of other previously published studies. To analyze QTc intervals in the study population, several categories were established based on available clinical evidence, particularly in terms of prognosis. The long QT categories used are those described by Goldenberg et al.,17 where the short QTc categories taken as a reference are those by Anttonen et al.18:

• •

  Normal QTc: 340-439ms

• •

  Borderline QTc: 440-469ms

• •

  Long QTc: ≥ 470ms17,19–21

• •

  Short QTc: 321-339ms

• •

  Very short QTc: ≤ 320ms11,18,22,23.

The analysis excluded individuals with complete left bundle-branch block or with paced or nonsinus rhythms.

Once the results were obtained, the new European Society of Cardiology guidelines on ventricular arrhythmia and sudden cardiac death,24 which establish a cutoff for long QT ≥ 480ms (class I, level of evidence C). Consequently, the study population data were also reanalyzed with this cutoff point.

Brugada 1 and 2 electrocardiographic patterns were defined according to the 2002 consensus document2:

• •

  Type 1 or coved pattern: convex-descending ST-segment elevation ≥ 2mm in more than 1 right precordial lead, followed by negative T-wave.

• •

  Type 2 or saddleback pattern: concave ST-segment elevation ≥ 2mm in more than 1 right precordial lead, followed by positive or isobiphasic T waves.

Because the sampling process meant that individuals in the population had differing probabilities of being selected, each participant in the final sample was assigned a weight that represented the number of persons from the Spanish population having the same sex, age bracket, and geographic area.25 Therefore, the sum of all weights of the sample was equal to the size of the Spanish population ≥ 40 years. This weighting process was complex and was undertaken in 2 phases, using the methodology described in previous reports on the OFRECE study13,14 combined with the procedure proposed by Deville and Särndal26 with the “Calibrate” instruction of the Stata statistical package (version 10.1). The 2011 municipal census of inhabitants by sex, age bracket, and geographic area was used as the population for the adjustment and calibration.

All analyses were performed in keeping with the sample design of the study. For all variables analyzed, specific prevalences were calculated by sex and age bracket along with the respective 95% confidence intervals (95%CI).

To identify the factors associated with the abnormalities of interest, odds ratios were estimated after adjustment for age and sex by logistic regression models. A multivariable model including factors with P < .1 in the previous analysis was subsequently fitted to identify factors independently associated with the conditions of interest.

Interobserver variability was also analyzed, and agreement was very high in all cases, above 99%. In terms of the consistency of Brugada-type patterns, expected agreement was 0.9947 and observed agreement was 0.9976. The Kappa index was 0.5444 (95%CI, 0.3667-0.7220). In the case of QTc, cases were noted and reviewed if the observer and reviewer measurements were more than 10ms apart (higher observer measurement, n = 172; higher reviewer measurement, n = 37). A third measurement was performed by both, and an agreement was reached. In all other cases (discrepancy ≤ 10ms), the average value was recorded in the database. These discrepancies in 209 (2.5%) pairs of measurements had a lower clinical impact because only 7 (0.08%) would have changed the patient classification in terms of QTc duration.

A total of 8343 participants completed the protocol, but only 7889 met the inclusion criteria for this analysis (native sinus rhythm and no left bundle-branch block). A total of 243 participants were excluded for AF, 135 for left bundle-branch block, 24 for both conditions, and 52 for nonsinus or unnatural rhythm. Of these, 52.5% were women with a mean age of 58.3 years. Most of the population (7011 individuals) had a normal QTc interval, thus accounting for 88.9% and for 90.5% when weighted by age and sex (95%CI, 89.5-91.5). In this group, mean QTc was 403.8ms (95%CI, 402.6-405.0). A total of 763 people had borderline QTc (prevalence of 9.7%; weighting, 8.3%; 95%CI, 7.4-9.3). Mean QTc was 449.3ms (95%CI, 448.5-450.0). A total of 96 participants had a long QTc interval (cutoff ≥ 470ms) (gross prevalence, 1.22%; weighted prevalence, 1.01%; 95%CI, 0.68-1.34). In this group, mean QTc was 484.9ms (95%CI, 480.2-489.6). Last, 51 participants had a QTc ≥ 480ms, with a prevalence of 0.65% and a weighted prevalence of 0.42% (95%CI, 0.26-0.57). Mean QTc was 499.48ms (95%CI, 493.11-505.85). A total of 18 patients had short QTc interval (mean QTc, 333.7ms; 95%CI, 331.6-335.9), for a prevalence of 0.23% and when weighted by age and sex, 0.18% (95%CI, 0.08-0.29). No patients had very short QTc interval. The characteristics of the population and of participants with QTc disorders are listed in Table 1. The multivariable analysis for QTc disorders (Table 2) showed that age and a history of AF were independently associated with the presence of long QT (cutoff ≥ 470ms). Hypertension was strongly associated in the univariable analysis. A multivariable analysis with cutoff ≥ 480ms also identified AF as an independent predictive factor of long QTc. The predictive factors for borderline QTc were age, female sex, hypertension, and diabetes mellitus. The association analysis excluded the patient group with short QTc due to its small size.

The Brugada ECG pattern analysis considered the data of 8343 individuals in the study, identifying 12 patients with Brugada 1 and 2 patterns:

• •

  Two patients had Brugada type 1 (2 women aged 55 and 54 years), giving a prevalence of 0.024%, or 0.01% when weighted (95%CI, 0.00-0.04).

• •

  Ten patients had Brugada type 2, yielding a prevalence of 0.12%, or 0.17% when weighted by age and sex (95%CI, 0.02-0.33). Eight of these were men; mean age of the group was 55.4 years.

Table 3 shows the clinical characteristics of participants with an electrocardiographic pattern consistent with Brugada 1 and 2 compared with those of the general population. The geographic distribution of patients is shown in Figure.

Figure.

Geographic distribution of the Brugada-type the electrocardiographic patterns found.

(0.31MB).

The prevalence found for Brugada 1 and 2 patterns was low (0.13%), a finding consistent with the European series published after 2002 in Greece, Italy, and Turkey.27–29 In northern countries (Germany, Finland, and Denmark),10,30,31 the prevalence was low and no Brugada type 1 pattern was observed. In our study, it is noteworthy that the 2 cases of Brugada type 1 corresponded to 2 middle-aged women, even though Brugada syndrome is traditionally considered to affect men more often, at a ratio of 8:1.2 In this regard, the findings of patients with Brugada type 2 patterns are more consistent with the medical literature (8 of the 10 patients reported were men). Our study determined prevalence only for individuals older than 40 years. However, in terms of prognosis, this selection bias may be of some importance because the above reviews report a mean age of 40 years for onset or diagnosis among affected patients.

The prevalence of QT abnormalities was high. According to Goldenberg et al.,17 borderline QTc or QTc above 470ms is associated with a higher risk of cardiac events. The prevalences found in this series are high: 8.3% and 1.01% (≥ 470ms), respectively. Therefore, we can estimate that almost 4 million people in Spain have some degree of QTc interval prolongation and are thus at risk of cardiac events. However, other studies have shown even higher prevalences. Veglio et al.20 reported a gross prevalence of 25.8% of people with QTc > 440ms. This cohort was small and selective: all participants had type 2 diabetes, mean age was 65.4 years, and most were women. Because QTc elongation is significantly related to age, female sex, and a diagnosis of diabetes, this alone explains why these authors observed higher prevalences than those obtained in the present study. The Rotterdam study21 also found much higher prevalences than those obtained in our study: 18% had borderline QTc and 11.1% had long QTc. Again, the mean age of the population analyzed was higher and there were more women than men. Furthermore, the cutoff points used were much more restrictive than those used in our study; consequently, all these factors may have contributed to the differences found between the 2 analyses. In our study, among affected patients, the long QT group showed a different clinical pattern than the group with borderline QTc. Borderline QTc identifies individuals with a high cardiovascular risk burden, whereas factors such as diabetes and hypertension are independent predictors. Cardiovascular risk burden had no influence on whether the patient had long QTc. In this case, QTc elongation above 470 or 480ms was strongly associated with prior episodes of AF. Regarding the potential role of antiarrhythmic drugs in this context, only 2 of the patients with a history of AF were receiving amiodarone or an HF class drug, and it was unknown how many participants were receiving these drugs for other arrhythmias; hence, it was not possible to accurately establish the percentage of individuals receiving antiarrhythmic therapy that might explain the presence of long QT. Regardless of any pharmacological effects, there may be unidentified characteristics in these individuals that cause QT elongation (eg, genetic factors related to congenital long QT syndrome) and be predisposing factors for AF. As in previous studies, borderline QTc was significantly related to age and female sex.17,20,21 This was not true of long QTc, which was related to age but not sex.

The prevalence of short QTc was 0.23% —somewhat lower than the prevalences published in Japanese series32,33 and more in line with European series11,18—,due to the cutoff used in this case (340ms or less).

Based on the results, more information is clearly needed on the true prognosis of these findings. A follow-up study in these populations will help identify patients who will benefit from more specific clinical follow-up.