Due to the nature of the sampling process, in which individuals within the population had different probabilities of being selected, each participant in the final sample was assigned a weight to reflect the number of people in the Spanish population of the same sex, and from the same age group and geographical area that the participant represented.14 Hence, the sum of all weights in the sample was equal to the size of the Spanish population aged 40 years and older. Weighting was complex and was done in 2 stages, according to the methods described in previous OFRECE studies.8,9 In the first stage, the weights to apply were defined, and in the second, the weights were calibrated according to the proposed procedure of Deville and Särndal15 using the calibrate instruction provided in the Stata v10.1 statistical software package. The 2011 municipal population census was used to adjust or calibrate the sample according to sex, age group, and geographic area.

All analyses took into account the sampling design of the study. The overall prevalence and specific prevalences by age and sex for all the variables analyzed were calculated with 95% confidence intervals (95%CI). When necessary, means or percentages were calculated, as appropriate, according to whether the variables were quantitative or qualitative.

To identify clinical factors associated with the ECG changes of interest, sex- and age-adjusted odds ratios (ORs) were estimated using logistic regression modeling. Subsequently, to identify factors independently associated with the conditions of interest, a multivariate model was constructed, including variables with p<0.1 in the previous bivariate analysis.

All the prevalence data presented in the results are weighted by sex and age (WP) with 95%CIs. Agreement between the first and second ECG reading was assessed using analysis of interobserver variance, applying Cohen's kappa coefficient to each of the variables analyzed.

This group of disorders was the most prevalent (WP, 17.5%; 95%CI, 16.3%-18.7%), and the results concur with those of Rodríguez-Capitán et al.2 As these authors also noted, conduction disorders generally showed a higher prevalence in men and tended to be more frequent with aging.

Right bundle branch block (complete or incomplete) was the most common conduction disorder (WP, 8.1%; 95%CI, 7.4%-8.9%). The prevalence of this condition was somewhat higher than the values reported in other series, perhaps because we included both complete and incomplete blocks in the analysis. On multivariate analysis there was a significant association of right bundle branch block with pulmonary disease, but not with heart disease.

Left anterior-superior hemiblock was the second most common conduction disorder found (WP, 6.5%; 95%CI, 5.9%-7.2%). There are no data on the prevalence of this abnormality in the older series.5–7 The KORA study3 reported a prevalence of 0.16% and Rodríguez-Capitán et al. 0.5%,2 likely because of the greater presence of younger individuals in both these studies. In our series, left anterior-superior hemiblock was associated with a higher cardiovascular risk burden on multivariate analysis, which suggests that this finding, in itself, should lead us to consider more exhaustive risk stratification in these patients.

Intraventricular conduction disorder is a nonspecific ECG abnormality that is often dismissed or included within right or left bundle branch blocks. We consider this an error, as the criteria regarding these 3 conduction disorders are different and well defined.12 In the classic series, there is no mention of the prevalence of this finding. Only KORA3 and the study by Rodríguez-Capitán et al.2 specifically analyze the prevalence of intraventricular conduction disorder, but no data are provided on related clinical factors. In our population, it was the third most common conduction disorder and its presence identified patients with a high cardiovascular disease burden. On multivariate analysis, it showed a strong association with hypertension, heart failure, and AF. Of note, the association between intraventricular conduction disorder and hypertension yielded an OR of 1.96, a value similar to that of left ventricular hypertrophy and hypertension (OR=2.07) and greater than that of left anterior-superior hemiblock and hypertension (OR=1.49). Hence, ECG findings of nonspecific intraventricular conduction disorder should be taken into account in the clinical evaluation of patients with suspected structural heart disease.

The prevalence of left bundle branch block in our series (WP, 1.6%; 95%CI, 1.1%-1.8%) was higher than the values reported by Ostrander et al. (0.35%)5 and Rodríguez-Capitán et al.,2 likely because of the lower mean age of the participants in these studies. Our prevalence is similar to the values reported by Rose et al.6 (between 0.98 and 3.18%, depending on age group) and De Bacquer et al.7 (between 0.1 and 2.0%, depending on sex and age). As expected, left bundle branch block identified patients with a larger cardiovascular disease burden (hypertension, heart failure), which is in keeping with the traditional perception that detection of this disorder, even in an asymptomatic patient, requires consultation with a cardiologist. Unlike the other conduction disorders, distribution by sex of left bundle branch block showed no significant differences.

As in other series, nonspecific repolarization changes were the most common ECG abnormalities (16% of participants). These findings generate diagnostic uncertainty and, because of their high prevalence, may imply numerous visits to the cardiologist and even emergency room admittance. Usually these findings are considered “nonspecific” (as the name indicates) and they may be given little importance. Nonetheless, a higher cardiovascular disease burden was found in this group of participants. Multivariate analysis associated this finding with AF and a greater coronary heart disease burden. Thus, nonspecific repolarization changes are not irrelevant findings; they may help to identify patients with higher cardiovascular risk.

Left ventricular hypertrophy was detected in 3.2% of the study participants. This percentage is similar to the values reported by Ostrander et al. (4%)5 and Rodríguez-Capitán et al. (3.6%),2 which is surprising as different diagnostic criteria were used. We consider that the findings of Rose et al.6 and De Bacquer et al.7 are not comparable to ours, as these authors refer to the prevalence of “tall R waves” and not specifically to left ventricular hypertrophy. Other series, such as the one in KORA3 showed a low prevalence of this abnormality (1.74%). Overall, a very low percentage of abnormal ECG findings was detected in KORA, perhaps because of the method used (automatic ECG analysis) and the characteristics of the sample. On multivariate analysis in the present study, hypertension was an independent predictor of left ventricular hypertrophy and surprisingly, a history of stroke was a protective factor. This is a striking finding that does not have a clear explanation. A higher prevalence of left ventricular hypertrophy would be expected in stroke patients as an expression of a higher cardiovascular risk burden (eg, greater hypertension). It could be that stroke patients undergo more intensive control of cardiovascular risk factors, such as hypertension, and this could lead to a smaller percentage of left ventricular hypertrophy.

The prevalence of first-degree atrioventricular block was 3.7%, similar to the values described in classic series,6 in which a higher prevalence of this abnormality was found in association with aging: 1.86% in the group aged 40 to 49 years and 3.39% in those aged 60 to 69 years. Although prolonged PR interval has been traditionally considered a benign finding, our results are more in agreement with those of Cheng et al.,16 who reported that prolonged PR is associated with AF, pacemaker implantation, and increased mortality. In our population, long PR was more prevalent in men (65.6% of the total with this finding) and was associated with a higher cardiovascular risk burden, which may indicate a need to change our view that prolonged PR is a minor finding. The very low prevalence of other, higher-grade atrioventricular blocks was surprising: There were only 3 Wenckebach blocks and no Mobitz II or complete blocks. In this line, Ostrander et al.,5 Rose et al.,6 and Rodríguez-Capitán et al.2 also found a very low prevalence of high-grade atrioventricular blocks.

Pathologic Q wave in an inferior location was present in 2% of participants, a somewhat higher value than has been reported in other series.2,5,6 The prevalence of pathologic Q wave in any location (anterior, inferior, lateral, etc.) in these reports ranges from 0.9% to 3.2%.2,5–7 As in other studies, inferior pathologic Q wave was the most prevalent and was associated with coronary artery disease (OR=7.64) and central obesity (OR=2.69). When inferior pathologic Q wave is detected, we should (of course) think of an underlying coronary substrate, but there is another possibility: a phenomenon related to horizontal positioning of the cardiac apex in persons with central obesity, in particular if the affected patient is a man.

Atrial extrasystoles were found in 2.4% of participants, a value higher than the 0.3% reported by Rodríguez-Capitán et al.,2 perhaps because our population sample was older. This abnormality was associated with a higher percentage of heart failure, which contradicts the traditional consideration that atrial extrasystoles are an ECG finding devoid of interest.

The prevalence of extrasystoles was 2.8%, a higher value than that reported by Rodríguez-Capitán et al.2 On multivariate analysis, no significant correlation was found with heart failure or other cardiac conditions. On the contrary, participants with ventricular extrasystoles showed a lower prevalence of coronary disease, which is difficult to explain, unless it results from greater use of beta-blockers in this group.

As in previous studies, there were very few examples of Wolf-Parkinson-White patterns in our population sample (prevalence, 0.35%). This value is comparable to the figures cited by De Bacquer et al.7 and Rodríguez-Capitán et al.,2 and somewhat higher than the prevalence reported by Rose et al.6 (0.02%).

The design of the study (cross-sectional with a population older than 40 years) implies that conclusions regarding the true prognosis of the ECG findings could not be obtained, and that information on individuals younger than 40 years was unavailable. In addition, it was not possible to verify that the electrocardiograms had been acquired according to proper quality standards (application of filters, adequate voltage and paper speed, etc.). These factors could not be controlled, and the analysis necessarily involved ECG studies performed with different electrocardiographs. This may have had an impact on some diagnoses, but a more rigorous design would have been complex and costly, threatening the feasibility of the study. An advantage of these technical considerations is that they rendered the study more similar to actual clinical practice. Some of these limitations were partly corrected by centralizing the electrocardiogram analysis. With regard to the ECG criteria used, it would have been of interest to differentiate between complete and incomplete bundle branch blocks. These were grouped together to simplify the huge amount of information managed in the study. For the same reason, we did not differentiate between Q wave duration greater or less than 0.04 s. Another important limitation is that complete data were not available for the pharmacological treatment that may have been used by participants, which could have had an impact on the appearance of certain ECG abnormalities: repolarization changes, conduction disorders, etc. The only information in this line was related to antiarrhythmic drug use in participants with AF.