Methods
Patients
The Mayo Clinic Institutional Review Board approved this cohort study. Consecutive patients aged ≥18 years who underwent transthoracic echocardiography between 1 January 2006 and 31 December 2011 with the following criteria were enrolled: (1) aortic valve area (AVA) <1.0 cm, (2) preserved LV EF (≥50%), (3) LG (mean gradient <40 mm Hg) and (4) absence of prosthetic valves, complex congenital heart disease, supravalvular or subvalvular AS, hypertrophic cardiomyopathy, and concomitant moderate or severe native valvular lesions. These criteria led to a final study population of 405 patients. The medical record was manually reviewed for symptoms, comorbidities and laboratory data.
2D and Doppler Echocardiography
Comprehensive 2D and Doppler echocardiographic studies were performed on commercially available ultrasound equipment (Acuson Sequoia, Siemens Medical, Mountain View, California, USA; Vivid-7, GE Healthcare, Milwaukee, Wisconsin, USA; and IE33, Phillips Healthcare, Andover, Massachusetts, USA) in accordance with the American Society of Echocardiography guidelines. Blood pressure was measured by manual sphygmomanometer and cardiac rhythm measured by electrocardiography at the time of echocardiography. EF was measured using the modified 2D Quinones formula or biplane method of disks. LV outflow tract diameter was measured in the parasternal long axis view in early systole from the point of aortic cusp insertion into the interventricular septum to the point of aortic cusp insertion into the intervalvular fibrosa. LV outflow tract time velocity integral was measured using pulsed wave Doppler by placing the sample volume just below the region of flow convergence at approximately 5 mm apically from the aortic valve in the apical long axis view and aligning it parallel with blood flow. LV SV was calculated by the Doppler method (LV outflow tract area×LV outflow tract velocity–time integral measured by pulsed-wave Doppler). An additional measurement of SV by the cube formula (SV=LVEDD−LVESD) was calculated for comparison with the Doppler method. A non-imaging probe was routinely used in multiple transducer positions (apical, suprasternal, supraclavicular, subcostal and right parasternal) in order to record the peak aortic jet velocity. The position that yielded the highest aortic valve velocity was used and at least three signals were traced and averaged to determine the time–velocity integral and calculate transvalvular pressure gradient. For patients in atrial fibrillation, 10 cardiac cycles were averaged to obtain the SV and mean aortic pressure gradient.
Afterload Assessment
Ventricular afterload was assessed using the methods derived from echocardiography and systolic blood pressure (10). Valvuloarterial impedance (Zva), a measure of global LV afterload, was calculated using the following formula: Zva (mm Hg/mL/m)=(mean systolic aortic valve Doppler gradient+systolic blood pressure)÷SVI. SAC, a measure of pulsatile arterial load, was measured using the formula: SAC (mL mm Hg m)=SVI÷(systolic—diastolic blood pressure). Systemic vascular resistance (SVR), a measure of non-pulsatile vascular load, was measured using the formula: SVR (dyne s cm)=80×mean blood pressure÷cardiac output.
Clinical Outcomes
Symptom onset, need for aortic valve intervention (valvuloplasty, transcatheter or surgical aortic valve replacement (AVR)) and vital status were determined using the medical record.
Statistical Analysis
Patients were stratified into quartile groups based on distribution of SVI. Additional comparisons were made with a group of patients with the most commonly used cutpoint for low flow used in the literature, SVI <35 mL/m. Data are reported as mean±SD or number and percentage for categorical variables. Student's t test was used to compare continuous variables and Fisher exact test to compare categorical variables between individual groups. Analysis of variance was used to compare multiple groups. Pairwise comparisons were performed using a post hoc Bonferroni significance level (p<0.008). Kaplan–Meier analysis with log-rank testing was used for temporal analysis of time to event outcomes in each group. An adjusted survival curve was also created using a semiparametric approach, assuming covariates follow the proportional hazards assumption while not requiring proportional hazards for SVI group. Survival of each group was compared with expected survival for an age and sex-matched Minnesota white population. The primary endpoint of interest was all-cause mortality. The secondary endpoint was all-cause mortality censored at the time of AVR. c-Statistics and the Akaike Information Criterion, a measure of model fit, were used to compare various SVI cutpoints for predicting overall mortality. A Cox proportional hazards multivariable model with stepwise elimination was used to determine predictors of all-cause mortality. Candidate variables included into the multivariable model included age, sex, body mass index (BMI), SVI, EF, AVA, mean gradient, peak velocity, Zva, SAC, right ventricular systolic pressure, hypertension, coronary artery disease, diabetes mellitus, atrial fibrillation, history of heart failure, prior transient ischaemic attack or stroke, chronic obstructive pulmonary disease, serum creatine, prior coronary artery bypass grafting surgery, and symptomatic status. c-Statistics were used to summarise the discriminatory ability of the new multivariable model compared with standard predictive variables. Difference in c-statistics and 95% CI were calculated and tested using the SE estimated from the 1000 bootstrap samples. The present study had approximately 80% power to detect a HR of 1.9 between two equal sized groups of 200 subjects. Statistical analysis was performed using SAS software V.9.3 and JMP software V.10.0, (Cary, North Carolina, USA). An a priori level of significance was determined at p<0.05.