Transvalvular jet velocity, aortic valve area, mortality, and cardiovascular outcomes Free

Abstract

Aims

The interplay between aortic stenosis (AS), cardiovascular events, and mortality is poorly understood. In addition, how echocardiographic indices compare for predicting outcomes remains unexplored for the full range of AS severity.

Methods and results

We prospectively calculated peak jet velocity (Vmax) and aortic valve area (AVA) in 5994 adult subjects with and without AS. We linked ultrasound data to 5-year mortality and clinical events obtained from electronic medical records. Proportional-hazard and negative binomial regression models were adjusted for relevant covariables such as age, sex, comorbidities, stroke-volume, LV ejection fraction, left valve regurgitation, aortic valve sclerosis or calcification, and valve replacement. We observed a strong linear relationship between Vmax and all-cause mortality (hazard ratio: 1.26, 95% confidence interval: 1.19–1.33 per 100 cm/s), cardiovascular events, as well as incidental and recurrent heart failure (HF). Adjusted risks were highly significant even at Vmax values in the range of 150–200 cm/s, risk curves separating very early after the index exam. Vmax was not associated with coronary, arrhythmic, cerebrovascular, or non-cardiovascular events. Although risks were confirmed when AVA was entered in place of Vmax, the risks estimated for categories based on the two indices were mismatched, even in patients with normal flow. An external cohort comprising 112 690 patients confirmed augmented risks of all-cause and cardiovascular mortality starting at values of Vmax and AVA in the range of mild AS.

Conclusions

Aortic stenosis is strongly associated to all-cause mortality, cardiovascular mortality, and cardiac events, specifically HF. Risks increase in parallel to the degree of outflow obstruction but are apparent very early in patients with mild disease. Criteria for grading AS based on Vmax and AVA are mismatched in terms of outcomes.

Open in new tabDownload slide

Introduction

Aortic valve stenosis (AS) has become a major public health problem and its prevalence is forecasted to double by 2050.¹ The survival of patients with AS is dismal when left untreated, but risks are substantially reduced by adequately timing aortic valve replacement (AVR). In fact, during the early phases of the disease, the prognostic impact of AS is classically assumed to be low.² Consequently, classical outcomes research in AS has aimed to balance the risks and benefits of AVR, focusing on the advanced phases of outflow obstruction.^3,4

This notion of a benign course of early AS has recently been challenged. Large retrospective registries have linked non-severe disease to all-cause^5–7 and cardiovascular-related⁵ mortality. Other cohorts point in the same direction.^8,9 However, the causes of such an excess mortality have been poorly clarified, and could be due to confounding factors.^10,11 Echocardiographic peak jet velocity (Vmax) and aortic valve area (AVA) are the ground of AS assessment and grading its severity relies on specific cut-offs of these two indices.¹² To date, how Vmax and AVA grading thresholds calibrate in terms of outcomes remains unknown, but this issue is particularly relevant because they are already known to be mismatched in terms of haemodynamic load.¹³

The present study was designed to clarify the association between AS and clinical outcomes in a contemporary cohort of patients (with and without a previous diagnosis of AS). We assessed the relationship between Vmax, AVA, and mortality, as well as cardiovascular and non-cardiovascular events. We then used risk curves to calibrate Vmax and AVA. Finally, we confirmed our major findings using outcome data from the National Echocardiography Database of Australia (NEDA)—a very large registry of unselected ultrasound studies.^5,14

Methods

The data underlying this article will be shared on reasonable request to the corresponding author.

Study design

From 1 July 2015 to 30 June 2016, we measured Vmax and AVA in all transthoracic echocardiograms performed in adult subjects at the Hospital General Universitario Gregorio Marañón of Madrid, Spain. This prospective measurement of Vmax and AVA in patients with and without AS was aimed to address the consequences of outflow obstruction without establishing any a priori threshold.¹⁵ The study was approved by the Institutional Ethics Committee, which exempted the need for informed consent and authorized for retrospectively retrieving baseline and follow-up data from electronic medical records (EMRs). Data were analysed anonymously, and the study was performed following current recommendations of the Declaration of Helsinki.

Image acquisition and analysis

Methodological details on the measurement of Vmax, AVA have been reported.¹⁵ Studies with prosthetic aortic valves or with subvalvular (late-peaking left ventricular outflow tract velocity ≥ 180 cm/s) or supravalvular obstruction (anatomical signs of supravalvular stenosis plus Vmax ≥ 200 cm/s) were excluded. We did not analyse the transvalvular pressure gradient data because it is tightly correlated with Vmax, and because the simplified Bernoulli equation is inaccurate if Vmax < 250 cm/s. We extracted the presence/absence of aortic valve sclerosis or calcification from the echocardiographic reports based on B-mode images.¹⁵ For patients with more than one echocardiographic exam, only data from the first one were analysed.

Clinical and outcome data

Medical history, comorbidities, and all data on admissions were obtained from EMRs based on the International Classification of Diseases ICD-9 and ICD-10 codes. Comorbidity was clustered using Charlson’s score. Baseline clinical and comorbidity data were obtained from all ICD codes of the EMRs available in 2 years before the index exam. An ‘event’ was defined as any unplanned hospital admission and was classified based on its primary diagnosis. Events were analysed as (i) the time from the echocardiographic study to the earliest event for each single or aggregated cause and (ii) the number of events (admissions) due to heart failure (HF). First HF events were tagged as incidental HF. Mortality data were obtained from the population information system of Madrid, Spain (CIBELES).

Study endpoints

The following primary endpoints were assessed for the 5-year follow-up period: (i) all-cause mortality, (ii) cardiovascular mortality (NEDA cohort only), (iii) any cardiovascular event (any event coded under ‘disease of the circulatory system’), (iv) HF events (both fatal and non-fatal), and (v) number of HF events (the number of hospitalizations divided by the follow-up period for each patient, expressed as the number of events every 2 years; see Results section). Secondary endpoints were (i) non-cardiovascular events (any event not coded as cardiovascular)—both aggregated and disaggregated for selected causes and (ii) other disaggregated cardiovascular events such as ischaemic heart disease, arrhythmia or stroke.

External cohort

The NEDA (Australian New Zealand Clinical Trials Registry ACTRN12617001387314) is a very large ongoing observational registry involving > 30 echocardiographic laboratories in Australia and represents a robust and exhaustive repository of outcome-matched clinical caseloads (Supplementary data online).^5,14,16

Statistical analysis

Data are presented as mean ± standard deviation except otherwise indicated. Univariable analyses for mortality (primary endpoint #1) and the HF admission rate (primary endpoint #5) were analysed by the Kaplan–Meier method and the Mantel–Haenszel ordinal association X² test, respectively. All primary endpoints were adjusted for covariables using either proportional-hazards (primary endpoints #1-4) or negative binomial (primary endpoint #5) regression models and calculated twice entering either Vmax or AVA as main predictors. Covariables were selected using clinical judgement, and the same set was entered in all multivariable models. Because missing clinical information at entry was frequent, we performed a sensitivity analysis using multiple imputation (Supplementary data online). In the NEDA cohort, Cox-proportional hazards models were constructed for all-cause mortality (primary endpoint #1) and cardiovascular mortality (primary endpoint #2) across the discrete Vmax and AVA groups. P values <0.05 were considered significant. Additional statistical details are presented in Supplementary data online.

Results

Study population, outcomes, and covariables

Complete baseline and follow-up [median (inter-quartile range): 4.5 (3.1–4.9) years] data were available for 5994 patients (22 270 person-years; Figure 1). The cohort was relatively old (mean age: 70 ± 18 years old, 56% male), with significant comorbidities (Charlson score: 3.6 ± 3.3). Study population characteristics are shown in Table 1. During the follow-up period, 833 (14%) patients were admitted for a cardiovascular event, and 101 patients underwent AVR (26 surgical and 75 percutaneous), at a mean time of 18 months after the index exam. A total of 2119 (35%) patients died. Relevant covariables related to mortality included age, male sex, inpatient condition during the index exam, diabetes mellitus, peripheral vascular disease, chronic renal disease, lower ejection fraction, calcific-degenerative changes of the aortic valve, and higher degrees of mitral and aortic regurgitation, amongst others (Table 2).

AS and outcomes

There was a linear relationship between Vmax and all-cause mortality (primary endpoint #1, Figure 2). Vmax accounted for a 26% excessive mortality for every 100 cm/s of Vmax (HR 95% CI: 1.19–1.33; Figure 2), evident already in the category of 150–200 cm/s. Vmax was also related to the risk of cardiovascular events (primary endpoint #3, Figure 3), mostly due to an increased risk of total and incidental HF events (primary endpoint #4). Remarkably, a 26% (HR 95% CI: 1.04–1.53) excessive risk for HF events was identified for patients with Vmax in the 150–200 cm/s range, which doubled in those in the 200–300 cm/s range. A direct relationship between Vmax and HF events was observed both in patients with and without LV systolic dysfunction (Figure 3). There was no association between Vmax and other cardiovascular or non-cardiovascular events (secondary endpoints). Actuarial analysis of incidental HF showed that all Vmax groups separated very early after the index exam (Figure 3). Vmax was related to the rate of hospital admissions (primary endpoint # 5), with an adjusted incidence rate ratio of 1.49 (95% CI 1.13–1.96) per every 100 cm/s (p < 0.001, Figure 4).

Outcome calibration of AVA and Vmax

The impact of AS on all primary endpoints was reproduced when AVA was entered in place of Vmax (Figures 2–4). However, this risk relationship was highly non-linear, remaining almost flat beyond 2.0 cm² (Figure 2). Importantly, adjusted risk equivalence for all primary endpoints was observed at smaller AVA values than expected from their Vmax counterparts (Figures 2 and 4). Consequently, the calibration curve showed risk equivalence at lower Vmax values than established in current grading criteria (Figure 5). As an example, the 42% adjusted mortality risk related to an AVA= 1.0 cm² (HR 95%, CI: 1.27–1.60) was reached at a Vmax value of 292 cm/s (HR: 1.42, 95% CI: 1.30–1.55). Survival mismatching of grading thresholds was highest in patients with low-flow but was also relevant in patients with normal-flow (Figure 5).

The sensitivity analysis yielded consistent risks of mortality and admission due to HF under different imputation assumptions (Supplementary data online).

External cohort

In the NEDA cohort, we confirmed significant relationships between AS severity, all-cause, and cardiovascular mortalities (endpoints #1 and #2, respectively, Figure 6). Again, even mild increases in Vmax or small reductions in AVA resulted in impaired outcomes, risks linearly increasing in parallel to severity. The risks related to Vmax and AVA cut-offs were also mismatched: the HRs of Vmax= 300–400 cm/s (1.86, 95% CI: 1.75–1.97 and 2.59, 95% CI: 2.46–2.96, for all-cause and cardiovascular mortality, respectively) were even higher than HRs of AVA values <1.0 cm² (1.69, 95% CI: 1.22–2.28, and 2.40, 95% CI: 2.17–2.66, respectively).

Discussion

This study demonstrates unequivocal relationships between AS, all-cause mortality, cardiovascular mortality, and cardiac events (particularly HF) which linearly grow parallel to the haemodynamic load of the disease. Excessive risks appeared even at mild degrees of valvular obstruction and remained significant when adjusted for confounders such as other left-heart valve diseases, stroke volume, ejection fraction, coronary artery and peripheral vascular disease, comorbidities, and even valvular degenerative changes.

‘Moderate’ AS has been related to excess all-cause and cardiovascular mortality in recent observational registries.^5,6,17 Because most laboratories do not calculate AVA in patients without a clear diagnosis of AS (e.g. Vmax> 200–250 cm/s), some selection bias in these studies is likely. Instead, in our cohort we calculated AVA prospectively in all consecutive patients referred for an exam, allowing us to calculate the dose–response risk relationship of AS for calibrating Vmax and AVA.

Given the observational nature of our study, we cannot fully address the mechanisms responsible for the observed impaired outcomes. Because we did not analyse sequential echocardiographic data, we cannot unambiguously rule out a role of haemodynamic progression; it can be argued that AS may have eventually become ‘severe’ before triggering clinical events. However, our actuarial analyses did not support this hypothesis. It takes in average 8 years to reach haemodynamic severity for patients with Vmax in the range of 100–200 cm/s, and usually >4 and 2 years for values in the mild and moderate range, respectively.^2,9,18 In fact, it has been recently shown that although ‘rapid progressors’ show reduced survival compared with ‘slow progressors’, actuarial curves are identical during the first year.¹⁹ However, in our study, we found that AS severity survival curves diverged in the first months after the index exam in two independent cohorts. Thus, although events in some patients may have been preceded by significant progression, it is more likely that AS impacts prognosis at milder degrees of haemodynamic burden than previously considered.

The role of other unmeasured confounders cannot be completely ruled-out either. Metabolic factors impact AS progression and may have contributed to the observed risks.²⁰ Coronary artery disease at least partially explains the adverse outcomes of patients with aortic sclerosis.¹⁰ To account for this, we adjusted prognostic models for peripheral and coronary artery disease, as well as for aortic valve sclerosis or calcification—a known imaging biomarker of coronary and noncoronary vascular disease. Inflammation has also been involved in the relationship between aortic sclerosis and cardiovascular outcomes.¹⁰ Indeed, inflammation is a well-known trigger of HF events and could participate of the observed associations.

Remarkably, we found a strong risk dose–response relationship of the degree of valvular obstruction, mimicking the well-established effects of systemic hypertension.²¹ Thus, we believe that the AS-related risk of HF may be due to the deleterious effects of the chronic pressure overload on the LV. Opposite to hypertension, AS was not associated to an increased risk of coronary or cerebrovascular events, probably because AS has a lower impact on the vascular compartment.²² The fact that AS was not related to an increased adjusted risk of coronary events supports that the potential interference caused by coronary atherosclerosis was adequately controlled for.

We identified an important prognostic mismatch of current Vmax and AVA severity thresholds. Patients with Vmax values of ‘moderate’ AS showed worse outcomes than those with ‘severe’ AS based on AVA. Ideally, severity indices of valvular heart disease should capture epidemiological (prevalence pattern in the general and risk populations), functional (relationship with the haemodynamic burden), and prognostic (correlation with outcomes) information. Because AVA normalizes for flow, it more accurately accounts for the functional severity of AS. However, non-flow corrected indices show better reproducibility¹⁵ and improved prognostic value.^3,23 Cut-offs for defining mild, moderate, and severe disease are 250, 300, and 400 cm/s and 2.0, 1.5, and 1.0 cm² for Vmax and AVA, respectively.¹² Because these values are obviously arbitrary, it should be desirable that each category accounts for a unique degree of functional impact and mortality risk. However, studies supporting current grading cut-offs were not designed to define best outcome stratification.^3,4,9,18 In fact, Vmax and AVA correlate modestly, and discordant grading is common. Under normal flow, a Vmax of 400 cm/s corresponds to an AVA closer to 0.75 than to 1.0 cm². Our results demonstrate that, in terms of outcomes, things seem to be similar.

Limitations

Many studies were unsuitable for analysis and clinical information was frequently unavailable, skewing our main cohort towards a mainly in-hospital population. However, the results of our sensitivity analysis and an external confirmation in a large independent cohort yielded very consistent HRs. Also, the rich case-mix of our study suggests that risk comparisons for Vmax and AVA are robust. Nevertheless, survival rates of our main cohort should not be projected to the general population in absolute terms. Particularly intriguing are the increased risks identified in the group of patients with Vmax= 150–200 cm/s. One third of these patients did not show valve sclerosis or calcification and increased risks persisted after adjusting for stroke volume (non-linearly), the presence of anaemia and other cofounding factors, but were absent in patients with AVA values in the 2.0–2.5 cm² range. Thus, the prognostic role of non-valvular factors capable of disturbing ejection fluid dynamics such as mild septal hypertrophy or impaired arterial impedance²² should be explored. We opted for conventional multivariable adjustment instead of the propensity score method to account for covariates.²⁴ The data of the NEDA cohort were collected retrospectively, and no imputation was performed for missing data. Thus, there may be some selection bias in this cohort. Other data acquisition limitations of our main cohort have been discussed elsewhere.¹⁵

Future directions

Our findings stress the need for a better understanding the biological bases of AS and its consequences in order to identify strategies for prevention. Drug research to slow progression must be encouraged and patients with very early disease must be identified.²⁵ Careful workup schemes must be widely deployed, and in light of our findings, current criteria used for establishing the diagnosis and grading AS may need to be revised. The potential incremental value of combining Vmax, AVA, and flow²⁶ should be investigated, as well as the added value of LV strain²⁷ or of integrated scores of cardiac damage.¹⁶

Importantly, a recommendation of anticipating AVR must not be inferred from our study. Any AVR procedure associates an inherent short-term risk as does living with a prosthetic valve over the longer term. Furthermore, the risks related to decades of myocardial remodelling caused by longstanding AS may not be fully reverted by AVR. Thus, only randomized multicentric clinical trials will clarify the impact of earlier intervention in patients with low Vmax and small AVA values.²⁸

Conclusions

Aortic stenosis is associated to impaired long-term clinical outcomes, following a strong risk relationship with all-cause mortality, cardiovascular mortality, cardiac events, as well as with incident and recurrent HF. The prognostic impact of AS remains stable after adjusting for multiple potential confounding factors. Excessive risks appear very early in patients with mild disease, suggesting that even low levels of outflow obstruction are clinically relevant. Current criteria for grading AS are mismatched in terms of prognosis, both in patients with normal and low transvalvular flow.

Supplementary data

Supplementary data are available at European Heart Journal—Cardiovascular Imaging online. 

Acknowledgements

We are in debt with all the personnel of the Laboratory of Cardiac Imaging from Hospital General Universitario Gregorio Marañón for their contribution to this study.

Funding

S.S. was supported by the NHMRC of Australia (GNT1135894). B.A. was supported by the Instituto de Salud Carlos III, Madrid Spain (INT19/00012).

Conflict of interest: GAS and DP are the co-principal investigators and directors of NEDA (a non-for-profit research entity). SS has received consultancy fees from NEDA. SS, DP and GAS have previously received consultancy/speaking fees from Edwards Lifesciences.

References

Author notes