OUP user menu

Prevalence, correlates, and prognostic significance of QRS prolongation in heart failure with reduced and preserved ejection fraction

Lars H. Lund, Juliane Jurga, Magnus Edner, Lina Benson, Ulf Dahlström, Cecilia Linde, Urban Alehagen
DOI: http://dx.doi.org/10.1093/eurheartj/ehs305 529-539 First published online: 4 October 2012

Abstract

Aims The independent clinical correlates and prognostic impact of QRS prolongation in heart failure (HF) with reduced and preserved ejection fraction (EF) are poorly understood. The rationale for cardiac resynchronization therapy (CRT) in preserved EF is unknown. The aim was to determine the prevalence of, correlates with, and prognostic impact of QRS prolongation in HF with reduced and preserved EF.

Methods and results We studied 25 171 patients (age 74.6 ± 12.0 years, 39.9% women) in the Swedish Heart Failure Registry. We assessed QRS width and 40 other clinically relevant variables. Correlates with QRS width were assessed with multivariable logistic regression, and the association between QRS width and all-cause mortality with multivariable Cox regression. Pre-specified subgroup analyses by EF were performed. Thirty-one per cent had QRS ≥120 ms. Strong predictors of QRS ≥120 ms were higher age, male gender, dilated cardiomyopathy, longer duration of HF, and lower EF. One-year survival was 77% in QRS ≥120 vs. 82% in QRS <120 ms, and 5-year survival was 42 vs. 51%, respectively (P < 0.001). The adjusted hazard ratio for all-cause mortality was 1.11 (95% confidence interval 1.04–1.18, P = 0.001) for QRS ≥120 vs. <120 ms. There was no interaction between QRS width and EF.

Conclusion QRS prolongation is associated with other markers of severity in HF but is also an independent risk factor for all-cause mortality. The risk associated with QRS prolongation may be similar regardless of EF. This provides a rationale for trials of CRT in HF with preserved EF.

  • Heart failure
  • Preserved ejection fraction
  • Reduced ejection fraction
  • QRS width
  • Epidemiology
  • Cardiac resynchronization therapy

Introduction

Heart failure (HF) affects 2–3% of the adult population and is associated with 50% mortality at 4 years, varying depending on severity.1 Heart failure with reduced ejection fraction (HFrEF), sometimes termed systolic HF, is well characterized and drug and device therapy improve prognosis.1 Heart failure with preserved ejection fraction (HFpEF), sometimes termed diastolic HF, is perhaps equally common and lethal as HFrEF,2,3 but is poorly characterized and there is no evidence-based therapy.

In the general population, QRS prolongation and/or left bundle branch block (LBBB) is present in <1% at middle age, increasing to 5–17% over age 80, occurring more commonly in male gender, HF, coronary artery disease, hypertension, and left ventricular hypertrophy, and associated with adverse outcomes.47

In HFrEF, QRS prolongation and/or LBBB is present in 24–47% and thought to be associated with increased mortality,811 although others report no increased risk.12,13 In HFpEF, non-specific QRS prolongation has not been studied but LBBB appears present in 8–40%, with a conflicting impact on outcome.9,14

These studies were relatively small, with limited covariate adjustment and potential confounding as well as conflicting results. Whether QRS prolongation is merely a risk marker of HF and its severity or actually a risk factor for HF progression and mortality is unknown. Reduced mortality with cardiac resynchronization therapy (CRT) in HFrEF1 suggests a risk factor effect of QRS prolongation that is responsive to intervention, but the independent clinical correlates and prognostic implications of QRS prolongation are still poorly understood. Whether CRT might be of benefit in HFpEF is unknown.

We hypothesized that QRS prolongation is an independent risk factor for mortality in both HFrEF and HFpEF, and that therefore CRT may be of benefit also in HFpEF. Therefore, we performed a comprehensive prospective assessment of the prevalence, correlates, and prognostic role of QRS prolongation in HFrEF and HFpEF.

Methods

Study protocol

The Swedish Heart Failure Registry (RiksSvikt) has been previously described.15,16 The inclusion criterion is clinician-judged HF. Eighty variables are recorded at discharge from hospital or after clinic visit and entered into an electronic database managed by the Uppsala Clinical Research Center (Uppsala, Sweden). The database is run against the Swedish death registry monthly. The protocol, registration form, and annual report are available at http://www.rikssvikt.se. Establishment of the registry and registration and analysis of data was approved by a multisite Ethics Committee. The registry and this study conform to the Declaration of Helsinki. Individual patient consent is not required or obtained, but patients are informed of being entered into national registries and are allowed to opt out.

Between 11 May 2000 and 12 March 2011, there were 57 879 registrations from 63 of 77 hospitals and 76 of 1011 primary care outpatient clinics in Sweden, representing 37 974 unique patients. QRS width is not recorded in patients with a pacemaker, CRT, and/or implantable cardioverter defibrillator (n = 5150), and of the remainder, 6933 were excluded because QRS width was missing, 9 because of missing follow-up, and 711 because of inconsistency between entered LBBB and QRS width (LBBB present and QRS <120 ms), leaving 25 171 patients.

Statistical analysis

Data handling and descriptive statistics were performed in SAS version 9.2 (SAS Institute, Inc., Cary, NC, USA) and all other analyses in R version 2.9.2 (R Foundation for Statistical Computing, Vienna, Austria). For all analyses, the level of significance was set to 5% and all reported P-values are two-sided.

QRS width was categorized by ≥120 vs. <120 ms and analysed as categorical unless otherwise stated. QRS width contains more physiological and statistical information as a continuous variable, but the dichotomization at the 120 ms cut-off is clinically more useful. Forty additional clinically relevant baseline variables (including LBBB) were selected for analysis (depicted in Table 1). To avoid bias and confounding due to baseline variables missing at random, multiple imputation (n = 10) was performed for all variables with any missing data, using predictive mean matching with QRS width as a continuous variable and variables in Table 1 (according to the aregImpute function in R: http://CRAN.R-project.org/package=Hmisc). The outcome, survival, was included as the Nelson–Aalen estimator in the imputation model, although not imputed itself since it contained no missing values. Descriptive statistics report raw data and specifies % missing for each variable, but all subsequent analyses were performed on imputed data with imputation corrections to the resulting standard errors. The same imputed values were used for all regression and subgroup analyses.

View this table:
Table 1

Baseline characteristics and odds ratios for associations with QRS ≥120 ms

VariableBaseline characteristicsOR for association with QRS ≥120 ms
Missing percentage of n = 25 171QRS <120; n = 17 478 (69%)QRS ≥120; n = 7693 (31%)UnivariableMultivariable
OR95% CIOR95% CI
Follow-up timea, days0.0712 (279–1258; 0–3723)643 (235–1167; 0–3723)
Deada0.06084 (35%)3320 (43%)
Clinical characteristics
 QRS width, ms0.094.8 (11.8)145.1 (19.3)
 Left bundle branch block#a7.80 (0%)4028 (56%)
 Age, OR per decadeb0.074 (12)76 (11)1.14***1.11–1.161.35***1.301.39
 Gender0.0
  Male9790 (56%)5347 (70%)1.79***1.69–1.891.65***1.551.77
  Female7688 (44%)2346 (30%)
 Civil status6.4
  Married/co-habitating8859 (54%)4215 (58%)1.19***1.12–1.251.061.00–1.13
  Single7501 (46%)2995 (42%)
 Living arrangements15.6
  Independent13 530 (92%)6088 (93%)1.090.99–1.211.110.99–1.24
  Institution1160 (8%)474 (7%)
 Location0.0
  Inpatient10 393 (59%)4588 (60%)1.010.95–1.061.050.98–1.12
  Outpatient7085 (41%)3105 (40%)
 Specialty7.4
  Internal medicine/geriatrics8387 (52%)3903 (54%)1.07*1.01–1.121.11***1.05–1.18
  Cardiology7687 (48%)3334 (46%)
 Year of inclusion0.0
  2000–053108 (18%)1545 (20%)1.16***1.09–1.241.030.95–1.11
  2006–1114 370 (82%)6148 (80%)
 Planned follow-up specialtya9.8
  Specialty care (cardiology or internal medicine)8665 (55%)3888 (56%)
  Other (other specialty, geriatrics)874 (6%)376 (6%)
  Primary care6238 (39%)2654 (38%)
 Planned follow-up nurse-based heart failure clinica9.65264 (33%)2427 (35%)
 Duration of heart failure, months0.2
  ≥67901 (45%)4264 (56%)1.51***1.43–1.591.37***1.29–1.46
  <69533 (55%)3415 (44%)
 New York Heart Association class29.6
  IV538 (5%)306 (6%)1.73***1.50–2.001.070.91–1.25
  III4140 (34%)2324 (41%)1.59***1.44–1.751.100.99–1.22
  II5838 (48%)2470 (44%)1.23***1.12–1.351.020.92–1.12
  I1574 (13%)519 (9%)
 Left ventricular ejection fraction, %15.4
  <303284 (22%)2754 (41%)4.10***3.77–4.453.98***3.61–4.38
  30–393906 (27%)1922 (29%)2.38***2.19–2.592.32***2.12–2.54
  40–493458 (24%)1178 (18%)1.64***1.50–1.791.63***1.48–1.79
  ≥503944 (27%)839 (12%)
Physical examination
 Systolic blood pressurec, mmHg1.6130 (22)128 (22)
 Diastolic blood pressurec, mmHg1.874 (13)72 (12)
 Mean arterial pressure, mmHg, OR per 10 unitsb1.893 (14)91 (14)0.91***0.89–0.920.96***0.93–0.98
 Pulse pressured, mmHg1.856 (18)55 (18)
  ≥50 mmHg11 128 (65%)4889 (64%)0.980.92–1.031.12***1.05–1.19
 Heart rated, b.p.m.9.576 (17)73 (16)
  >1001097 (7%)336 (5%)0.61***0.54–0.690.68***0.60–0.78
  75–1006544 (41%)2479 (36%)0.76***0.72–0.800.77***0.72–0.82
  <758194 (52%)4129 (59%)
Laboratory
 Creatinine clearanced, mL/min7.866 (36)61 (32)
  <50 mL/min6217 (39%)3125 (44%)1.20***1.14–1.270.93*0.86–1.00
 Haemoglobin, g/L, OR per 10 units0.0132 (18)132 (17)1.011.00–1.031.02*1.00–1.04
 NT-proBNPd, ng/L74.84562 (6871)5992 (8378)
 Log NT-proBNP, ng/L7.6 (1.4)7.9 (1.3)1.17***1.15–1.191.06***1.03–1.09
Comorbidities
 Smoking24.1
  Yes1868 (14%)710 (12%)0.920.84–1.000.930.85–1.03
  Previously5569 (42%)2633 (45%)1.09**1.03–1.160.960.90–1.02
  Never5828 (44%)2492 (43%)
 Hypertension2.98146 (48%)3449 (46%)0.94*0.89–0.991.040.98–1.10
 Diabetes mellitus0.6
  Insulin-treated1373 (8%)683 (9%)1.17**1.06–1.291.080.97–1.20
  Orally treated1170 (7%)573 (7%)1.15**1.04–1.281.110.99–1.24
  Both insulin and orally treated613 (3%)279 (4%)1.070.93–1.241.160.99–1.36
  Diet-treated915 (5%)464 (6%)1.19**1.06–1.341.100.97–1.25
  No13 305 (77%)5647 (74%)
 Ischaemic heart disease3.87860 (47%)4031 (55%)1.37***1.30–1.441.050.98–1.13
 Dilated cardiomyopathy4.41485 (9%)1091 (15%)1.80***1.65–1.951.60***1.45–1.77
 Hypertrophic cardiomyopathy17.8236 (2%)116 (2%)1.090.88–1.331.170.94–1.45
 Valve disease5.13152 (19%)1635 (22%)1.23***1.16–1.321.09*1.01–1.17
 Atrial fibrillation/flutter0.48243 (47%)3209 (42%)0.80***0.76–0.850.79***0.74–0.85
 Lung disease2.63248 (19%)1320 (18%)0.91**0.85–0.980.950.88–1.03
Drug therapy
 β-Blocker0.614 421 (83%)6314 (82%)0.970.90–1.040.83***0.76–0.89
 ACE-inhibitor0.610 876 (63%)4927 (64%)1.08**1.02–1.141.17***1.08–1.26
 Angiotensin receptor blocker2.03121 (18%)1605 (21%)1.21***1.14–1.301.34***1.22–1.47
 Aldosterone antagonist0.74642 (27%)2387 (31%)1.25***1.18–1.321.08*1.01–1.15
 Diuretic0.413 767 (79%)6363 (83%)1.29***1.21–1.391.09*1.01–1.18
 Digoxin0.63362 (19%)1250 (16%)0.82***0.76–0.880.87**0.81–0.95
 Nitrates0.72850 (16%)1563 (20%)1.31***1.22–1.401.14**1.05–1.23
 Aspirin and/or other platelet inhibitor0.69220 (53%)4350 (57%)1.17***1.10–1.231.000.92–1.08
 Oral anticoagulants0.75857 (34%)2438 (32%)0.92**0.87–0.980.950.87–1.03
 Statin0.57211 (41%)3309 (43%)1.08**1.02–1.140.970.90–1.03
Interventions
 History of PCI and/or CABG1.43908 (23%)1979 (26%)1.20***1.13–1.270.970.89–1.05
 History of valve surgery/intervention0.5700 (4%)510 (7%)1.71***1.52–1.921.92***1.68–2.20
  • Data are mean (standard deviation) or n (%), with the exception of follow-up, which lists median (inter-quartile range, range).

  • Boldface indicates values statistically significant in multivariable analysis.

  • P-values are nominal, without correction for multiplicity. With Bonferroni adjustment (correcting for 37 tests), the following variables were no longer statistically significantly associated with QRS prolongation: creatinine clearance, haemoglobin, aldosterone antagonist, diuretic.

  • OR, odds ratio; when comparing two or more groups, the reference is the group with no OR listed; CI, confidence interval; NT-proBNP, N-terminal brain natriuretic peptide; PCI, percutaneous coronary intervention; CABG, coronary artery bypass graft.

  • aNot included in logistic regressions (left bundle branch block not present in narrow QRS; death and follow-up time, planned follow-up specialty and planned follow-up in heart failure nurse clinic considered future events which may affect prognosis but should not affect present association with QRS width).

  • bAge and mean arterial pressure are entered in the model as continuous but ORs and 95% CIs are presented here in units of 10 for clinical interpretability.

  • cUsed to derive mean arterial pressure and pulse pressure. Not included in imputations or regressions, but included here for completeness.

  • dPresented as continuous but because of non-linear association with QRS width, categorized or log-transformed for logistic regressions, as appropriate.

  • *P < 0.05.

  • **P < 0.01.

  • ***P < 0.001.

Association of baseline variables with QRS width

Several univariable and multivariable logistic regressions were performed with QRS width as the outcome (Table 1). The le Cessie–van Houwlingen goodness-of-fit test was performed for the multivariable model and the resulting P-value was low, which could indicate problems with the overall fit of the model; however, this could also be an effect of the large number of observations and since no discrepancies between the predicted and observed values were detected when validating the model using bootstrapping (200 replications; http://CRAN.R-project.org/package=Design), the latter was deemed probable. The c-statistic was 0.70, considered acceptable discrimination.17 No extreme outliers were detected (Cook's distance <1 for all observations) and the largest variance inflation factor value was 3, so multicollinearity was deemed low. Linearity for the continuous variables was assessed using smoothed partial residuals plots and as a result N-terminal pro-brain natriuretic peptide (NT-proBNP) was log-transformed, whereas pulse pressure, heart rate, and creatinine clearance were categorized according to clinically relevant cut-offs (Table 1).

Survival according to QRS width

Survival for QRS width was charted with the Kaplan–Meier method. Several univariable and multivariable Cox regressions, using the Efron method for tie handling, were performed separately for QRS width as a categorical and continuous variable (Table 2). In order to evaluate the individual effect of the respective variables on the association between QRS width and mortality, bivariable models were also estimated for QRS width. The proportional hazards assumption was investigated for the scaled Schoenfeld residuals from the multivariable models and the dfbetas [a measure of how much the hazard ratio (HR) changes due to the deletion of a single observation] from the models were inspected to detect outliers,18 but no problems were detected. The assumption of linearity for the continuous variables was investigated using smoothed Martingale residuals plots, and variables deemed not to have a linear relationship with the outcome, creatinine clearance and NT-proBNP, were modelled using restricted cubic splines (enabling non-linear relationships between these variables and the outcome to be modelled) with 4 degrees of freedom.18

View this table:
Table 2

Hazard ratios for death associated with QRS prolongation, for the overall population and four ejection fraction subgroups

QRS categorical, ≥120 ms vs. <120 ms
ModelHR95% CIP-value
Univariable1.321.26–1.38<0.001
Adjusted fora
  Age1.221.17–1.28<0.001
  Gender1.351.30–1.41<0.001
  EF1.371.31–1.43<0.001
  NYHA class1.241.19–1.30<0.001
  NT-proBNP1.171.12–1.22<0.001
  Dilated cardiomyopathy1.371.31–1.43<0.001
  Creatinine clearance1.251.20–1.31<0.001
  Duration of HF1.271.22–1.33<0.001
Multivariable complete1.111.04–1.18<0.001
Subgroups, multivariable complete
 EF <30%1.171.03–1.320.017
 EF 30–39%1.221.07–1.400.003
 EF 40–49%1.100.94–1.290.225
 EF ≥50%1.060.90–1.240.479
QRS continuousb, per 10 ms
Model
Univariable1.041.03–1.05<0.001
Multivariable complete1.031.01–1.04<0.001
Subgroups, multivariable complete
 EF <30%1.021.00–1.040.059
 EF 30–39%1.041.02–1.070.001
 EF 40–49%1.041.01–1.070.008
 EF ≥50%1.031.00–1.050.046
LBBB or non-LBBB, multivariable complete
 LBBB vs. QRS <120 ms1.111.05–1.18<0.001
 Non-LBBB, QRS ≥120 ms vs. QRS <120 ms1.111.04–1.18<0.001
  • Abbreviations are as in Table 1. EF, ejection fraction; LBBB, left bundle branch block.

  • aThese variables, when adjusted for in bivariable analysis, had the largest impact on the HR for QRS width and/or had notable and relevant associations with QRS width.

  • bQRS was included as a continuous variable in the Cox regression but HRs and 95% CIs are presented for units of 10 ms for clinical interpretability.

Patient inclusion occurred largely before NT-proBNP was widely used, and 74.8% of patients had NT-proBNP missing. Therefore, a consistency analysis was performed excluding NT-proBNP from the multiple imputation and Cox regressions.

Interactions between QRS width and all covariates were estimated within the multivariable Cox regressions. Continuous variables were analysed as such when assessing interactions but dichotomized at clinically relevant cut-offs when displayed in a forest plot (Figure 3). Four pre-specified and clinically relevant subgroup Cox analyses were performed, by EF <30, 30–39, 40–49, and ≥50%.

Results

Baseline characteristics and associations with QRS width

Baseline characteristics are shown in Table 1 and schematically in Figure 1. Of the 25 171 patients (age 74.6 ± 12.0 years, 39.9% women), 31% had a QRS width of ≥120 ms. QRS width ≥120 ms was present in 46, 33, 25, and 18% in the EF groups <30, 30–39, 40–49, and ≥50%, respectively (P < 0.001). Numerous characteristics were independently associated with QRS ≥120 ms (Table 1). Notable were odds ratio (OR) of 1.65 for male gender, 1.60 for dilated cardiomyopathy, 1.06 for log NT-proBNP, and 3.98, 2.32 and 1.63 for EF <30, 30–39, and 40–49%, respectively, compared with ≥50%.

Figure 1

QRS width distribution. (A) Histogram of QRS width (ms); (B) QRS width ≥120 vs. <120 ms by ejection fraction, New York Heart Association class, and gender. As ejection fraction increases, the proportion of women increases, New York Heart Association class improves, and the proportion with QRS ≥120 ms decreases. Data are not imputed.

Outcome

One-year survival was 77% in QRS ≥120 vs. 82% in QRS <120 ms, and 5-year survival was 42 and 51%, respectively (log-rank P < 0.001) (Figure 2). Table 2 lists HRs for death for QRS width as a categorical (QRS ≥120 vs. <120 ms) and continuous variable (per 10 ms increase). For QRS ≥120, the univariable HR was 1.32 [95% confidence interval (CI) 1.26–1.38, P < 0.001] and the multivariable HR was 1.11 (95% CI 1.04–1.18, P < 0.001). In bivariable analysis, NT-proBNP, NYHA class, age, creatinine clearance, and duration of HF were the single variables that, when adjusted for alone, considerably decreased the HR of QRS prolongation.

Figure 2

Kaplan–Meier survival by QRS width. The difference in survival was significant (P < 0.001). The difference in survival remained significant after adjustment for 40 covariates (Table 2).

Clinically relevant interactions are shown in Figure 3. There were several statistically significant interactions between QRS width and covariates; however, in none of the resulting subgroups did the entire 95% CI of the HR fall below 1.00. The interactions between QRS width and NYHA class and QRS width and EF were not significant, but there were signals towards greater risk associated with QRS prolongation with lower EF and with better NYHA class.

Figure 3

Multivariable hazard ratios for all-cause mortality by subgroups. P-values are for interaction between QRS width and the variable on the y-axis. Numbers are average of 10 imputations. NYHA, New York Heart Association; LVEF, left ventricular ejection fraction.

In a sensitivity analysis without NT-proBNP, the multivariable HR for QRS ≥120 was 1.14 (95% CI 1.08–1.21, P < 0.001).

Of the 31% with QRS ≥120, 56% had LBBB. The multivariable HR for LBBB vs. QRS <120 ms was 1.11 (95% CI 1.05–1.18, P < 0.001) and for non-specific intraventricular conduction delay (QRS ≥120 without LBBB) vs. QRS <120 ms, the multivariable HR was also 1.11 (95% 1.04–1.18, P < 0.001) (Table 2).

Subgroup outcomes

Table 2 depicts HRs by QRS width for the four EF subgroups. QRS prolongation was independently associated with increased mortality as a categorical and continuous variable for all subgroups with a few exceptions (Table 2). QRS prolongation was associated with higher excess hazard in lower EF subgroups. However, as mentioned, there was no interaction between EF and QRS width as a continuous (P = 0.610) or categorical (P = 0.256, Figure 3) variable.

Discussion

QRS prolongation on the electrocardiogram represents intraventricular conduction delay, and since it was first noted to be associated with increased mortality in HF in the 1960s,19 it has been thought to be merely a risk marker of underlying structural heart disease. However, with the success of CRT in the 2000s,1 there is renewed interest for the clinical correlates of QRS prolongation and its role as a direct risk factor for HF progression and death, amenable to intervention. We show that QRS prolongation ≥120 ms was present in 31% of patients with HF in the Swedish Heart Failure Registry. It was more common with reduced than with preserved EF, but once it occurred, it was an independent risk factor for mortality regardless of EF, suggesting CRT may be beneficial also in HFpEF. There was a signal towards greater risk with better NYHA class, suggesting the potential benefit of CRT could actually be greater in milder HF.

Correlates with and risk factors for QRS prolongation

QRS prolongation occurs in the general population, increasing sharply with age.46 Normal ageing may be associated with fibrosis in the conduction system.20 QRS duration is also related to cardiac size and thus gender, and is associated with hypertension, left ventricular hypertrophy and fibrosis, and ischaemic heart disease, all of which may be subclinical.46 In our study, each decade of life was associated with a 35% increased odds of QRS ≥120 ms, independently of other covariates known to be associated with QRS prolongation and severity of HF, suggesting that ageing itself is important not only in the general population but also in HF.

QRS prolongation is mainly associated with structural heart disease, such as in HF. In our study, QRS ≥120 ms was present in 31% overall and in 39% of patients with reduced EF (<40%), 25% in patients with mildly reduced EF (40–49%) and 18% with normal EF (≥50%). These prevalences were slightly higher than previous reports of QRS prolongation/LBBB in HFrEF812 and slightly lower than the few previous reports of LBBB in HFpEF,9,14 and they confirm the important proportional and independent association of LV dilatation and systolic dysfunction with QRS prolongation.21 The ORs of QRS prolongation associated with EF was similar in univariable and multivariable analyses, suggesting EF is not only a risk marker but also a risk factor for QRS prolongation. However, the 18% prevalence of QRS prolongation in the EF ≥50% group was still higher than in the general population,46 suggesting that the pathophysiology of HFpEF (cellular and ventricular hypertrophy, fibrosis of the conduction system and myocardium, ischaemia, and diastolic dysfunction22) contributes to QRS prolongation despite normal EF.

Dilated cardiomyopathy was strongly associated with QRS prolongation, independent of EF, and the ORs were similar in multivariable and univariable analyses, again suggesting a risk factor role. In contrast, ischaemic aetiology was not associated with QRS width, suggesting that ischaemic aetiology and focal fibrosis or scar are less important than the global diffuse fibrosis and the increased distance for impulse propagation seen in LV dilation. In two studies,23,24 QRS was wider in more advanced NYHA class but multivariable analysis was not performed. In our study, NYHA class was associated with QRS prolongation in univariable but not multivariable analysis, suggesting NYHA class is a marker for other factors affecting QRS width, such as severity of HF, but not in itself a risk factor for QRS prolongation.

Male gender is associated with wider QRS in the general population,6,25 and here we show that male gender increased the odds of QRS prolongation in HF by 65%. This was independent of all covariates, suggesting that larger cardiac size is important. QRS width ≥120 ms is the cut-off for CRT (in NYHA III–IV), regardless of gender,1 but recent data suggest that women may respond better to CRT,26,27 suggesting that in women, the severity of electromechanical dyssynchrony may be greater for a given QRS width.

Heart failure is associated with progressive remodelling and small studies suggest that QRS progressively prolongs in dilated cardiomyopathy28 and after myocardial infarction.29 In our study, duration of HF ≥6 months compared with new onset (<6 months) was associated with a 37% increase in the odds of having QRS ≥120 ms. This was independent of other markers of progression, such as EF, suggesting that progressive remodelling in the conduction system itself contributes.

Prognostic impact of QRS prolongation

In the general population, QRS prolongation may predict future need of pacemaker,4 cardiovascular mortality,30 and possibly but not conclusively6,31 all-cause mortality. In HFrEF, most810 but not all11,12 studies suggest that QRS prolongation/LBBB is associated with increased cardiovascular or all-cause mortality. These studies are of varying size and with variable adjustment for confounders. Here we show that in HF overall, QRS ≥120 ms was associated with 32% increased hazard for death in univariable analysis and 11% increased hazard after adjustment for 40 covariates. The hazard was progressive and linear, with a 4% increase per 10 ms in univariable analysis and 3% in multivariable analysis. The difference in magnitude in uni- and multivariable analyses suggests that QRS prolongation is a marker of other risk factors and severity of HF, but the presence of an independent hazard suggest that it is also a causative risk factor.

Prognostic impact of QRS prolongation in milder heart failure

NYHA class was not independently associated with QRS prolongation, but once present, QRS prolongation could be associated with different risk depending on NYHA class. Although the interaction between QRS width and NYHA class was not significant, there was a signal towards greater risk with better NYHA class. Thus, in mild HF, the relative importance of QRS prolongation and thus the potential benefit of CRT may be greater, adding credence to recent studies of CRT in mild HF.32,33 In these studies, entry criteria were QRS ≥120 and 130 ms, respectively, but the recent European Society of Cardiology guidelines recommend a QRS ≥150 ms for CRT in milder HF.34 If QRS prolongation can be confirmed to be more harmful in milder HF, perhaps this indication should be expanded.

Left bundle branch block

Most studies report data on QRS prolongation4,10,11 or BBB/LBBB,5,6,8,9 but not both, and the relative prevalence and prognostic importance of LBBB are poorly understood, although LBBB appears more responsive to CRT than non-specific intraventricular conduction delay and particularly, RBBB.35 In our study, QRS prolongation was harmful, independent of LBBB, and when analysed separately, both LBBB and non-specific intraventricular conduction delay were similarly harmful.

Prognostic impact of QRS prolongation in heart failure with preserved ejection fraction

The role of QRS prolongation in HFpEF is poorly studied. Non-specific QRS prolongation has not been studied, but in small studies, LBBB was present in 8–40%, with conflicting impact on outcome.9,14 We show that QRS width predicted mortality independent of EF and that there was no interaction between EF and QRS width, although there was a signal towards lesser risk with preserved EF. Overall, QRS prolongation was harmful whether analysed as a categorical variable (most useful for clinical decision-making and with regard to CRT indication) or as a continuous variable (containing more detailed physiological information). In pre-specified subgroup analysis, EF 40–49 and ≥50% did not reach statistical significance for QRS ≥120 vs. <120 ms, but was significant and of equal magnitude for QRS width as a continuous variable.

Taken together, these data suggest that QRS prolongation is an independent risk factor for all-cause mortality also in HFpEF, and set the stage for potential trials of CRT in HFpEF. Interestingly, patients included in PROSPECT but with a core laboratory EF >35% may have derived similar benefit from CRT as those with lower EF.36

Limitations

Our study has limitations. QRS width was entered at local medical centres and was not subject to adjudication. Indeed, there are no firm criteria for establishing QRS width.37 The registry contains data on QRS width and LBBB but does not separate non-specific intraventricular conduction delay vs. RBBB. N-terminal pro-brain natriuretic peptide was not in widespread use until the later years of the study and was missing in and imputed for 74.8% of patients. To avoid potential confounding due to excessive imputation, rather than exclude NT-proBNP from our analyses altogether, we performed a separate sensitivity analysis without NT-proBNP, which altered the HR only from 1.11 to 1.14, both P < 0.001. We do not have data on subsequent hospitalization or device implantation.

QRS prolongation was an independent predictor of mortality after adjustment for a large number of covariates known to predict prognosis, suggesting but not proving it is not merely a risk marker but also a risk factor for mortality. However, we cannot rule out that unknown and/or unmeasured confounders may explain some of this risk.

Conclusion

We show that QRS prolongation occurs in HF and is independently associated most strongly not only with dilated cardiomyopathy and lower EF but also with age, male gender, and duration of HF. Notably, it is not independently associated with severity of disease and thus may occur equally often in milder disease. QRS prolongation and LBBB are also independent risk factors for all-cause mortality, also in preserved EF and milder HF. Our findings should be validated in independent prospective studies.

The independent association of QRS prolongation and LBBB with mortality suggests that they are risk factors amenable to intervention, and provides a rationale for prospective trials of CRT in HFpEF and potentially for expanded indications for CRT in milder HF.

Funding

The Swedish Heart Failure Registry is funded by the Swedish National Board of Health and Welfare, the Swedish Association of Local Authorities and Regions, the Swedish Society of Cardiology, and the Swedish Heart-Lung Foundation. This work was supported by the Swedish Heart Lung Foundation (grants 20080409 and 20100419 to L.H.L.) and the Stockholm County Council (grant 00556–2009 to L.H.L.). No funding agency had any role in the design and conduct of the study, in the collection, management, analysis, or interpretation of the data, or in the preparation, review, or approval of the manuscript.

Conflict of interest: There are no commercial products involved in this study. However, if this study in any way affects future studies of or use of cardiac resynchronization therapy, there may be a perceived conflict of interest with the pacemaker industry. Therefore, we disclose the following: L.H.L.: research grants to author's institution from Medtronic, Inc. C.L.: research grants to author's institution from Medtronic, Inc.; principal investigator of REVERSE, a CRT study sponsored by Medtronic, Inc. Other authors: none declared.

Acknowledgements

We thank the local investigators at the 63 hospitals and 76 primary care outpatient clinics reporting to the Swedish Heart Failure Registry for this study.

References

View Abstract