Methods

This is a retrospective cohort study on prospectively collected data designed to assess the relation between QRS duration on routine ECG and CV mortality in a US population using NHANES III. NHANES III was conducted by the National Center for Health Statistics and is a cross-sectional study that includes data from oral surveys and general health examinations designed to assess demographic, socioeconomic, dietary, and overall health status in a nationally representative sample of noninstitutionalized patients from all 50 states. In NHANES III every subject aged >40 years old underwent baseline 12-lead electrocardiography at rest in a supine position. The ECG was recorded using a Marquette MAC 12 (Marquette Medical Systems, Inc., Milwaukee, Wisconsin), which uses simultaneous multiple leads, thus facilitating the recording of superior quality ECGs. The mobile examination center physician was responsible for approving the quality of the ECG based on predefined minimum standards. ECGs were then analyzed at the Epidemiological Cardiology Research Center at Wake Forest University using the NOVACODE (Wake Forest University School Of Medicine, Winston-Salem, North Carolina) electrocardiographic program, which classified ECGs according to the Minnesota Coding System.

Our initial cohort of 8,561 subjects was selected from all adults enrolled in NHANES-III from 1988 to 1994 with available electrocardiographic data. We then excluded patients with missing QRS duration (n = 30) and mortality data (n = 4). Our study population consisted of 8,527 participants from the NHANES-III with available complete ECG data. This cohort is representative of 74,062,796 individuals in the United States. There was a follow-up period of 12.5 ± 4.5 years (Mean ± SD) per patient and 106,244.6 person-years. Multiple imputation was used to circumvent missing data.

Hypertension was defined as self-reported history or use of antihypertensive medication(s) or systolic blood pressure >140 mm Hg and/or diastolic blood pressure >90 mm Hg on physical examination. Both systolic and diastolic blood pressures were calculated from an arithmetic mean of 6 or fewer readings. Similarly, hypercholesterolemia was defined as self-reported history or use of lipid-lowering medication or low-density lipoprotein cholesterol values above the normal cutoff as per the National Cholesterol Education Program Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) guidelines. Family history of heart attack was defined as participants having any first-degree relative with this diagnosis or who died from a heart attack before 50 years of age. CAD was defined as a participant’s self-reported history of myocardial infarction/heart attack, patient history of typical angina pectoris, or ECG changes suggestive of myocardial infarction (probable or possible myocardial infarction by the Minnesota Code). Heart failure was defined as a participant’s self-reported history of heart failure or receiving treatment for heart failure. Diabetes mellitus was defined as a participant’s self-reported history of diabetes, use of diabetes medication, or glycosylated hemoglobin ≥6.5%. The Modification of Diet in Renal Disease formula was used to calculate glomerular filtrate rate. Serum potassium is presented as milliequivalents per liter. The normalized calcium value was derived from adjusting the measured ionized calcium for pH. Intraventricular conduction delay was defined according to Minnesota coding system as QRS duration ≥120 milliseconds in a majority of beats in any of leads I, II, III, aVL, or aVF. LBBB was defined as QRS duration ≥120 milliseconds in a majority of beats in any of leads I, II, III, aVL, and aVF plus R peak duration ≥60 milliseconds in a majority of beats (of the same QRS pattern) in any of leads I, II, aVL, V5, V6. RBBB was defined as QRS duration ≥120 milliseconds in a majority of beats in any of leads I, II, III, aVL, and aVF, plus R’ > R in V1 or V2; or QRS mainly upright, with R peak duration ≥60 milliseconds in V1 or V2; or S duration > R duration in all beats in lead I or II.

Mortality status was obtained from the NHANES III–linked mortality public-use file, which provided mortality data through December 31, 2006. Underlying causes of death were provided by death certificate data contained in the same mortality files and classified according to International Classification of Diseases, Tenth Revision injury and cause-of-death guidelines. Codes for CV disease were identified according to American Heart Association guidelines and included codes I00 to I99 for circulatory disease.

NHANES-III had a complex nonrandom multistage stratified sample design. Analyses were performed using the designated weighting specified in the NHANES-III data set to minimize biases. We used the total NHANES-III pseudo-stratum as our strata variable, the total NHANES-III pseudo-primary sampling units as our survey sampling units, and the total mobile exam center final weight as our sampling unit weight. We then stratified our study sample by quartiles of QRS duration as follows: Q1, 61 to 89 milliseconds; Q2, 90 to 97 milliseconds; Q3, 98 to 105 milliseconds; and Q4, 106 to 209 milliseconds. Differences in baseline characteristics were examined using 1-way analysis of variance for continuous variables (reported as mean ± SD) and chi-square test for categorical variables (reported as a %).

The association between QRS duration and study outcomes over time was examined using Cox proportional hazards modeling. We adjusted for potential confounding in a model that included all the assessed covariates with the exception of low-density lipoprotein cholesterol due to its high colinearity with other lipid variables. We created 3 multivariate models: Model 1, based on quartiles of QRS duration; Model 2, based on QRS morphology (LBBB, RBBB, or intraventricular conduction delay); and Model 3, 10-millisecond incremental increase on QRS duration. All statistical analyses were conducted in STATA SE 11.1 (StataCorp, College Station, Texas). A 2-sided p value <0.05 was considered significant.

For reclassification, we created 2 models: Model A, covariates of the Framingham Risk Score (FRS); and Model B, covariates of the FRS and QRS duration in 10-millisecond increments. Because Model B was a “special case” of Model A, we used the likelihood ratio test to assess the differences in global measure of model fit. The Bayes Information Criterion was calculated to evaluate the improvement in the global measure of model fit after addition of QRS duration to Model A.

The Bayes Information Criterion takes into account the number of variables in the predictive model and prefers a model with fewer variables if it provides equally good prediction. A lower value implies better fit. We then calculated predicted 10-year CV risk estimates based on both models for all study participants and directly compared them with the actual risk observed during follow-up period. To calculate the net reclassification improvement, we divided the data on the basis of 10-year CV mortality risk into the following categories: <5%, 5% to 10%, 10% to 20%, and >20%. A score of 1 was assigned to every correct reclassification, which implies every individual who experienced an event or nonevent was upgraded or downgraded in the risk category respectively. A score of −1 was assigned to every incorrect classification based on the aforementioned criteria. A score of 0 was assigned to every “nonreclassification.”

We also calculated the integrated discrimination index (IDI), which along with net reclassification improvement offers useful insights into model improvement. IDI is the measure of improvement imparted by the addition of QRS duration to the FRS model. Absolute IDI is calculated as (Ň new,events − Ň new,nonevents) − (Ň old,events − Ň old,nonevents), where Ň new,events and Ň new,nonevents are the means of the new model-based predicted probabilities of events and nonevents, respectively, and Ň old,events and Ň old,nonevents are the means of the old model-based predicted probabilities of events and nonevents, respectively. Relative IDI is an index calculated when the incidence of events is relatively small. Relative IDI is expressed as [(Ň new,events − Ň new,nonevents)/(Ň old,events − Ň old,nonevents)] − 1.

The distributions of baseline demographic, laboratory, and ECG characteristics by quartiles of QRS duration are shown in Tables 1 and 2 . Our data set consisted of 8,527 subjects, 53% women and 87% classified as white, with mean age of 60.5 ± 13.6 years. During the study period, 1,433 CV deaths occurred. On multivariable analysis age, male gender, hypertension, current smoking, history of myocardial infarction in family or self, and history of heart failure, diabetes, and stroke were independent significant predictors of CV mortality ( Supplementary Tables 1, 2, and 3 ). Significant predictors of increasing QRS duration were male gender, white race, body mass index, history of heart failure, serum potassium level, and left ventricular hypertrophy ( Supplementary Table 4 ).

Table 1

Baseline characteristics of study population according to quartiles of QRS interval

Variable	All	Q1 (61–89)	Q2 (90–97)	Q3 (98–105)	Q4 (106–209)	p Value for Trend
	n = 8,527	n = 2,154	n = 2,329	n = 2,050	n = 1,994
Age (yrs)	60.5 ± 13.6	61.0 ± 13.9	59.9 ± 13.4	57.0 ± 13.3	62.2 ± 13.6	0.8
Male gender	47.2	26.4	40.0	54.6	70.5	<0.01
White race	87.1	84.8	86.3	88.1	89.2	0.03
Body mass index (kg/m 2 )	27.7 ± 5.5	27.8 ± 5.7	27.5 ± 5.6	27.8 ± 5.3	28.1 ± 5.5	<0.01
Hypertension	43.7	40.8	42.8	43.2	48.5	0.02
Smokers	23.6	23.9	22.4	22.1	26.5	0.1
Hypercholesterolemia	59.7	59.3	60.7	58.0	60.9	0.6
CAD	15.8	16.4	14.3	13.6	19.8	0.003
Heart failure	3.1	2.7	2.4	2.8	4.7	0.004
Previous stroke	2.9	2.9	3.2	1.8	3.9	0.03
Diabetes mellitus	11.2	11.3	12.0	9.3	12.1	0.08
Family history of heart attack	11.2	10.8	11.6	11.4	10.7	0.89
Serum total cholesterol (mg/dl)	218.0 ± 44.1	220.4 ± 45.8	218.9 ± 44.5	216.5 ± 42.1	216.1 ± 43.9	0.004
Serum high-density lipoprotein (mg/dl)	50.8 ± 16.3	53.3 ± 16.7	51.4 ± 16.3	50.4 ± 16.2	48.0 ± 15.4	<0.01
Serum triglycerides (mg/dl)	162.8 ± 130.2	158.1 ± 132.4	160.6 ± 115.0	160.9 ± 145.0	172.3 ± 128.0	0.003
Calculated serum low-density lipoprotein (mg/dl)	136.3 ± 38.8	136.8 ± 39.5	136.6 ± 39.1	136.1 ± 37.7	135.3 ± 38.6	0.58
Glomerular filtration rate (ml/min)						0.07
≥90	5.43	4.61	5.00	7.03	5.11
60–90	67.15	64.61	68.22	67.97	67.61
30–60	27.01	30.39	26.45	24.59	26.78
<30	0.4	0.39	0.33	0.41	0.5
Serum normalized calcium (mmol/L)	1.2 ± 0.1	1.2 ± 0.1	1.2 ± 0.1	1.2 ± 0.1	1.2 ± 0.1	<0.01
Serum potassium (mEq/L)	4.1 ± 0.4	4.1 ± 0.4	4.1 ± 0.4	4.1 ± 0.4	4.1 ± 0.4	0.0001
Major ECG abnormalities	14.1	8.2	6.8	9.0	35.4	<0.01
Minor ECG abnormalities	18.3	16.2	18.6	20.5	17.7	0.11
Heart rate (beats/min)	68.7 ± 11.8	70.8 ± 12.2	69.1 ± 11.9	67.6 ± 11.3	66.9 ± 11.4	<0.01
QT interval (ms)	407.7 ± 32.2	398.1 ± 30.9	404.8 ± 30.8	410.1 ± 31.4	419.3 ± 32.4	<0.01
Corrected QT interval (ms)	432.2 ± 24.5	428.8 ± 23.4	430.5 ± 23.5	432.0 ± 24.0	438.4 ± 26.2	<0.01
PR interval	164.9 ± 29.1	164.2 ± 27.6	162.5 ± 28.1	164.0 ± 28.7	169.3 ± 31.7	<0.01
LVH	13.2	10.0	11.4	10.3	22.1	<0.01

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Usefulness of Intravascular Ultrasound to Predict Outcomes in Short-Length Lesions Treated With Drug-Eluting Stents
2. Effectiveness of Moxonidine to Reduce Atrial Fibrillation Burden in Hypertensive Patients
3. Usefulness of Severe Cardiac Sympathetic Dysfunction to Predict the Occurrence of Rapid Atrial Fibrillation in Patients With Wolff-Parkinson-White Syndrome
4. Effectiveness of Spironolactone Plus Ambrisentan for Treatment of Pulmonary Arterial Hypertension (from the [ARIES] Study 1 and 2 Trials)

Stay updated, free articles. Join our Telegram channel

Join