Meta‐Analysis of P‐Wave Dispersion Values in Healthy Individuals: The Influence of Clinical Characteristics

Abstract

Background: P‐wave dispersion (Pd) is an appealing marker for predicting the risk of developing atrial fibrillation. At present, no definitive cutoff value has been determined as to the diagnosis of high‐risk patients. Our aims were to evaluate P‐wave parameters of healthy subjects published in the literature, determine normal range and weighted means of Pd and P‐wave parameters, and investigate the influences of gender, age, and BMI on the weighted results.

Methods: A systematic search of studies published in PubMed was conducted. Only studies which included control groups of healthy individuals were included.

Results: Of the 657 studies initially identified, 80 were eligible for inclusion. The total number of participants was 6,827. The highest reported Pd values were 58.56 ± 16.24  ms; the lowest were 7 ± 2.7  ms. The weighted mean was 33.46 ± 9.65  ms; weighted median was 32.2  ms. Gender and age were not found to be associated with significant influences on P‐wave parameter values. High‐normal BMI was not found to be associated with increased P‐wave parameter values.

Conclusions: Pd, Pmax, and Pmin span a wide range of values in healthy individuals. Seemingly, abnormal values were often reported in healthy adults. The high variability of P‐wave parameters in healthy individuals, and overlapping of the results with those reported for patients with increased risk for atrial fibrillation, might suggest that this technique has limited sensitivity and specificity. The variability between studies may stem from methodological issues and, therefore, there is a definite need for methodological standardization of Pd measurements.

Ann Noninvasive Electrocardiol 2012;17(1):28–35

INTRODUCTION

Atrial fibrillation (AF) is the most common form of arrhythmia causing severe medical complications, as well as early death. ¹ AF may be encountered in patients with cardiac disease, although in others no underlying clinical condition was found. The traditional risk factors for AF involvement are advanced age, male gender, heart failure, valvular disease, and hypertension. ² Scientific attention has been focused on attempting to identify patients with an increased risk for AF evolvement.

P‐wave dispersion (Pd) is calculated by subtracting the minimal P wave (Pmin) duration from the maximal P wave (Pmax) duration, as measured by multiple surface ECG leads, from a single beat. Measurement of Pd is based on an assumption that surface ECG represents partially regional changes in myocardial activation. Nevertheless, it remains unknown whether Pd is determined only by heterogeneity of atrial conduction or by other factors as well. ³ , ⁴ Pd values of >40  ms were found to be correlated with AF, with a sensitivity of 74–83% and specificity of 81–85%. ⁵

Nevertheless, a consensus regarding Pd values, distinguishing normal patients from high‐risk patients, is lacking. Various research groups have included control subjects with different clinical characteristics. Cardiovascular risk factors, like obesity, hypertension, coronary artery disease, valvular disease, and diastolic dysfunction, were all suggested to influence the Pd values. Nonetheless, the influences of demographic parameters such as gender, age, and body mass index (BMI) on P‐wave parameters remain to be determined. Therefore, our aim was to conduct a meta‐analysis of P wave parameters in healthy study groups, in an attempt to estimate their normal range.

METHODS

The meta‐analysis was conducted in accordance with the “preferred reporting items for systematic reviews and meta‐analyses.” ⁶

Eligibility Criteria

Only articles published in English were included. The control group was completely healthy. Control group studies which included patients with diabetes, hypertension, coronary artery disease or other congenital or acquired heart condition, lung diseases, collagen vascular diseases, thyroid dysfunction, rhythms other than sinus, or use of any type of medication were excluded. Studies with an insufficient amount of data regarding the health status of the participants were also excluded from analysis. Patients who underwent coronary angiography prior to the study, due to suspected coronary disease, regardless of the results were excluded, as well as patients with a history of coronary artery disease associated with morbidity. Patients with a history of supraventricular or ventricular arrhythmias, obstructive sleep apnea, or alcoholism were also excluded. Smoking was not an exclusion criterion. Control patients enrolled prior to surgery were also excluded.

Information Source

A PubMed‐based search was conducted on Jan 15, 2011.

Search Strategy

The following key words were used “P‐wave dispersion” (202 results), “P dispersion” (36 results), “p‐disp” (5 results), “dispersion AND P[Title] AND wave[Title]” (185 results), ““P wave” AND dispersion” (262 results). Overall, 657 unsorted results were initially found.

Data Synthesis

Studies were selected in which Pd was calculated according to the following equation: Pd equals Pmax minus Pmin. When a study included two healthy groups of patients who were exposed to influences not affecting Pd (such as high altitude), ⁷ both groups were included in the analysis. Studies where measurements were repeated several times a day, morning values were chosen for the analysis. However, it is noteworthy that values in the controls were similar throughout the day. ⁸ Also, when healthy individuals were evaluated for Pd on several occasions (i.e., during different seasons), ⁹ yearly results were averaged. Only studies, where mean ± SD values were presented, were included. In cases when more than one technique was employed for P‐wave measurements, only manual measurements were included. Healthy athletes were also referred to as controls. In cases where both averaged beats and nonaveraged beats were calculated, the randomly chosen beat was used.

Statistical Analysis

Results were expressed as mean and standard deviation (SD). Correlation analyses were performed using the linear, logarithmic, and quadratic regressions. The number of patients was used as a weighted variable. A probability value of P <0.05 was considered significant; two‐tailed P values were used for all statistics. Normal distribution was evaluated with the Shapiro–Wilk test. Analyses were performed using SPSS 15 for Windows software (SPSS, Chicago, IL, USA) and JMP version 7.0 (SAS Institute, Cary, NC, USA). The cutoff for normal values was determined according to 95% confidence interval.

RESULTS

After the exclusion of references traced in more than one search strategy and exclusion of five irrelevant papers, 251 references were included in the initial screening; 103 references did not report any control group and were therefore excluded from the analysis. Eight articles were excluded due to unorthodox techniques of Pd calculations (i.e., evaluation of Pd from Magnetocardiography, when dispersion was calculated as SD of P‐wave durations or calculation of Pd from 16 precordial leads). Eleven were excluded since they were not written in English. Three articles were excluded due to insufficient data on the health status of the participants; 38 were excluded due to the inclusion of a control group of unhealthy individuals. Two articles were excluded due to the presentation of results in medians and not means. Overall 80 articles were included in the meta‐analysis (Fig. 1). ^{3 , 7} , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁸ , ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² , ⁶³ , ⁶⁴ , ⁶⁵ , ⁶⁶ , ⁶⁷ , ⁶⁸ , ⁶⁹ , ⁷⁰ , ⁷¹ , ⁷² , ⁷³ , ⁷⁴ , ⁷⁵ , ⁷⁶ , ⁷⁷ , ⁷⁸ , ⁷⁹ , ⁸⁰ , ⁸¹ , ⁸² , ⁸³ , ⁸⁴ , ⁸⁵ Fourteen of these articles utilized only a computer‐assisted technique for P‐wave evaluation.

Figure 1.

Methods for inclusion of articles in the study.

The meta‐analysis results of mean P wave parameters were of normal distribution. Pd results were available in 80 studies, which included 86 eligible study groups. One paper reported on Pmax, Pmin, and mean P wave duration but not on Pd. The highest reported Pd values were 58.56 ± 16.24  ms (the patients were healthy athletes, N = 810 participants). ⁸⁴ The highest value for normal non‐athletic patients was 55 ± 15  ms (N = 37). ¹⁴ The lowest Pd values were 7 ± 2.7  ms (N = 70), ⁵⁸ and 7 ± 2.9  ms (N = 62). ³⁰ The weighted mean was 33.46 ± 9.65  ms; weighted median was 32.2  ms. The biggest cohort, which included 1,353 patients, reported on mean Pd values of 38.36 ± 10.38  ms. ³⁹   Figure 2 demonstrates the association between logarithmic presentation of the number of included participants (logN), and Pd values. The scattered lines represent the funnel‐like shaped range of the 99% confidence interval of expected Pd values, according to the biggest cohort. ³⁹ Twenty‐nine groups only were found to be in the expected zone (33.7% of the study groups). Thirteen studies reported mean Pd values of >40  ms (15.1% of all the study groups). Pmax was evaluated in 75 studies. The highest reported Pmax values were 125 ± 15  ms; ³¹ the lower were 64 ± 9.4  ms. ²⁸ Weighted mean was 99.3 ± 11.5  ms; the weighted median was 99.4  ms. Pmin was reported in 62 studies. The highest reported Pmin values were 104 ± 9.2  ms. ¹⁰ The lowest reported Pmin values were 31.1 ± 5.4  ms. ⁸² Weighted mean Pmin was 65.3 ± 13.0  ms; weighted median was 66  ms.

Figure 2.

Pd values compared with logarithmic scale of number of participants (N). The funnel‐like scattered lines represent the expected 99% confidence interval of Pd values according to the biggest scaled study. Only 29 groups were found in the expected zone (33.7% of the study groups).

No association was found between gender and P‐wave parameters. In addition, no significant association was found between Pmax and Pmin, and BMI. Nevertheless, a significant negative association was found between Pd and BMI according to all models (Fig. 3, R square 0.47–0.48, P < 0.001). A separate analysis of studies which included more than 30 participants, and which have specified BMI values (18 study groups, including 1646 participants), was associated with an even stronger association (R square ∼0.58, P < 0.001). According to the calculated 95% confidence interval, the upper cutoff was 54.8  ms for Pd, 124.8  ms for Pmax, and 97.3 for Pmin. The upper normal values for Pd in patients with BMI values lower than 25  kg/m² was 58.56  ms, and for patients with BMI values higher than 25  kg/m² was 42.5  ms.

Figure 3.

Bivariate fit of mean Pd by BMI according to linear, logarithmic, and quadratic models. A significant negative association was found between Pd and BMI consistent with all models (R square 0.47–0.48, P < 0.001).

No correlation, by any regression analysis, was found between age and Pd, Pmax and Pmin values (R square ∼0.1). A separate analysis of studies which included more than 30 individuals (49 study groups) did not alter the results.

DISCUSSION

In the present study, P‐wave parameters were evaluated in healthy individuals. Although 40  ms was proposed as a value distinguishing patients with a high risk for developing AF from low risk patients, ³ , ⁴ , ⁵ we have found that these high values were commonly reported in healthy individuals as well. Therefore, it seems that in Pd (as was previously observed in QT dispersion measurement) overlapping of the results between normal and abnormal conditions occurred. ⁸⁶ We suggest that small sample size may by associated with inappropriately high or low Pd values, compared with expected values, according to big cohorts.

We did not find that a higher male percentage was associated with higher Pd values and higher Pmax. This association has been investigated in only a few studies. Kose et al. did not observe gender differences in children. ⁵⁵ Sari et al. reported similar findings in a small group of healthy males and females. ⁶⁵ In contrast, Yildiz et al. included a very large cohort and reported that male gender was associated with higher Pd and Pmax values. ⁸⁴ Sari et al. and Kose et al. results are consistent with the findings in our meta‐analysis. It should be emphasized that although in general AF is more common in males, ² none of the included investigated groups reported that their male control patients ultimately developed AF.

The negative associations found between increased BMI and Pd is intriguing, since clinically an opposite association was reported by most groups. In a recently published review by Rosiak et al., it was reported that an increase of a single BMI unit was associated with an increase of about 8% in the risk for developing AF. ² Also, higher Pd was reported in obese patients compared with patients with normal weight in studies specifically aimed at evaluating such an association. ⁶⁸ , ⁸⁷ A negative association between BMI and Pd, consistent with the findings of the meta‐analysis, was also reported by Magnani et al. after applying a multivariable adjusted model (P = 0.04), ⁸⁸ further supporting our results. Nevertheless, mean BMI values were in the current study within the normal‐to‐overweight range and none of the included studies had patients with mean BMI values higher than 28  kg/m². Similarly, Magnani et al. included patients with a BMI lower than 30  kg/m². ⁸⁸ Therefore, it is reasonable to assume that our conclusions should be applied only to patients with a BMI lower than 28  kg/m². We speculated that the association between BMI and Pd was inverted in BMI values above this range. Therefore, we hypothesized that different meta‐analysis or a large prospective cohort design including only obese patients with higher BMIs might yield a different association. Furthermore, it is possible that the studies which included thinner patients had unrecognizable conditions predisposing them to increased risk for supraventricular arrhythmias, hence higher Pd values. Interestingly, the weighted upper normal values of Pd in patients with mild overweight (i.e., 42.5  ms) were similar to those suggested as the upper normal values which distinguish high‐risk patients for AF development (i.e. approximately 40  ms). In some studies which suggested a Pd cutoff, the BMI values were not reported. ¹⁵ , ¹⁹ , ³³ A different study reported on both Pd and BMI. Nevertheless, despite BMI values lower than 25 kg/m², their Pd suggested cutoff for identifying high risk patients was not increased. ⁸⁹

We did not find any significant association between increased age and Pd values. Although the risk of AF increased dramatically in older patients, ⁹⁰ in the majority of the studies, mean age was 25–45 years, in which range AF is less frequent. Yildiz et al. also found no association between age and Pd. ⁸⁴ Magnani et al., in contrast, reported a positive association between age and all P‐wave parameters. ⁸⁸

The major limitation of the present study was the inclusion of heterogeneous cohorts in diverse research groups utilizing different P‐wave measurement methodology. For instance, accuracy and reproducibility may significantly increase when PC‐based on‐screen measurements are applied. ¹⁵ Studies may differ in applying manual versus automated measurements, ECG on paper compared with digitalized ECG, paper speed and printed resolutions, ECG sampling rate, the number of investigators who evaluated P‐wave parameters, the number of included leads, etc. A comparison of different studies, applying measurement techniques, due to the lack of measurement standardization, is challenging. Also, P‐wave parameters and Pd may vary due to external influences such as seasonal effects. ⁹

Internal influences such as anxiety may also dramatically alter the results, ⁷⁷ and since anxiety levels were not routinely measured, statistical standardization could not be performed. Another internal factor that may have altered the results was whether the healthy patients received adequate sleep. It was found that following sleep deprivation in young adults, Pmin significantly decreases, Pmax increases and Pd increases. ⁶⁵ Pd may also be influenced by autonomic function, ⁴ thus no standardization could be performed in the current study design. Diurnal variation may also pose as an intervening factor since most studies did not elaborate on the time of the day in which ECG measurements were conducted. ⁸ It should also be emphasized that there was a vast difference in the number of included patients among studies. The mean number of patients was 79.4 ± 184.2 (median 34.0, 75% quartile 52.0, 25% quartile 24.8 patients). Therefore, by definition, studies which included larger cohorts had greater weighted influence on the results.

CONCLUSIONS

Pd, Pmax, and Pmin span a wide range of values in healthy individuals. Gender and age have no significant effects on P‐wave parameters. Higher BMI within the nonobese range is associated with lower Pd values. Seemingly abnormal values of Pd were commonly reported in healthy adults. The high variability of P‐wave parameters in healthy individuals, and overlapping of the results with those reported for patients with increased risk for AF, might suggest that this technique has limited sensitivity and specificity. Since Pd varies with BMI, it should be taken into account when evaluating whether a study group is within the normal range. Moreover, the current studies emphasized the need for Pd measurement standardization. It is quite reasonable to assume that such standardization will increase the reproducibility, sensitivity, and specificity of this test.

Study Limitations

In the current study design, the raw data of all included studies were unavailable for analysis. Therefore, we conducted the meta‐analysis by evaluating weighted means according to the accepted method. We cannot predict if an analysis of the raw data would have yielded different results, and therefore, the interpretation of our results should be done within this context.