Impact of Age on Aortic Valve Calcium Progression and Risk for Aortic Stenosis: Multi-Ethnic Study of Atherosclerosis

Introduction

Calcific aortic valve disease (CAVD) is the most common valvular disease in the United States ¹ with a prevalence that is expected to increase by up to three-fold over the next 25 years². Currently, the only available treatment is aortic valve replacement (AVR), which has very high costs ^{3, 4}. However, without AVR, patients face a poor prognosis with a 15% five-year survival ⁵ and even with AVR, most patients already have irreversible myocardial damage, which has been associated with worse survival ^{6, 7}. The poor prognosis associated with advanced CAVD emphasizes the need for better risk stratification and early detection to identify persons at higher long-term risk, who may be more likely to benefit from novel treatment approaches.

The prevalence of CAVD is higher among older persons^8–14 due to a higher prevalence of CAVD risk factors and longer exposure time to these risk factors ^{10, 15, 16}. CAVD, which is fundamentally characterized by aortic valve calcification (AVC), has a prevalence as high as 55% among participants ≥80 years ⁸. AVC can be easily measured using cardiac computed tomography (CT), which is commonly used to measure coronary artery calcification (CAC). Greater AVC values are associated with a higher long-term incidence of severe aortic stenosis (AS), with AVC ≥300 associated with an over than 300-fold increase in the long-term risk for severe AS ¹⁴.

While previous research has highlighted a higher prevalence of AVC >0 among older populations and a very strong association between AVC >0 and AS, limited data exists on 1) whether there is a difference in the rate of AVC progression for younger versus older persons or 2) if the association between AVC and moderate-severe AS differs by age. Therefore, in this study we examined the progression of AVC and the association of AVC with long-term risk of moderate-severe AS for younger versus older persons in the Multi-Ethnic Study of Atherosclerosis (MESA).

Methods

Study Population

A total of 6,810 MESA participants between 45 and 84 years old who were free of known CVD were included. Participants were from six different sites throughout the United States (Baltimore, New York, Chicago, Los Angeles, Forsyth County, and St. Paul). The MESA design has been previously described ¹⁷. The Institutional Review Board reviewed and approved the MESA protocol at each institution, and written consent was obtained from all participants. For this study, participants were excluded if they were missing a baseline AVC score (n=2) or had severe AS at baseline (n=2).

At MESA Visit 1 (2000-02), data regarding age, sex, race/ethnicity, income, education, current smoking status, medical history, and current medications were self-reported by participants. Diabetes was defined as use of current anti-glycemic medication, fasting glucose ≥126 mg/dL, or self-report. Seated blood pressure was recorded a total of three times, and the average of the final two readings was used for analysis. LDL-C was calculated using the Friedewald equation.

Assessment of Aortic Valve Calcification and Aortic Valve Calcification Progression

AVC was measured using non-contrast cardiac CT at MESA Visit 1 (2000-02). ECG-gated cardiac computed tomography scanners were used in the Chicago, Los Angeles, and New York Field Centers while a four-slice multidetector computed tomography system was used in the Baltimore, Forsyth County, and St. Paul Field Centers. AVC was scored for calcified lesions in the aortic valve leaflets including those that extended into the aortic root, but not for those that were isolated to the aortic ring ^{17, 18}. AVC was quantified using the Agatston method¹⁹. All scans were read centrally at the Harbor-UCLA Research and Education Institute core reading lab. AVC measured using cardiac CT has low intrareader and interscan variability, 4.4% and 9.7% respectively¹⁸.

A total of 911 participants at AVC >0 at MESA Visit 1. Follow up CT scans were performed among participants with baseline AVC>0 as follows: 25% of MESA participants were randomly selected to have follow up CT scans conducted at MESA Visit 2 (2002-04) and Visit 3 (2004-05), and 50% of the participants were randomly selected to have a follow up CT scan conducted at MESA Visit 5 (2010-11). A total of 499 participants with AVC >0 at Visit 1 had a follow-up CT scan to measure AVC progression and were included in the AVC progression analysis. The same protocol for measuring and quantifying AVC was utilized for baseline and all follow up scans.

Assessment of Incident Moderate-Severe Aortic Stenosis

Participants and/or families were contacted every 9 to 12 months and reported all hospital admissions, outpatient CVD diagnoses, or deaths. Incident moderate-severe AS was adjudicated as has previously been described ^{14, 20, 21}. In brief, medical records were then obtained for corresponding admission, outpatient diagnoses, or death. ICD codes were used to identify candidate moderate-severe AS events, and if present, complete medical records were obtained. Two Cardiologists from the MESA CVD Events Adjudication Committee independently reviewed the medical records, and any disagreements on event classification were discussed until a consensus was obtained. This was supplemented by MESA Visit 6 echocardiography data. Echocardiography was performed for 3,032 of 3,303 (91.8%) of participants attending exam 6 (2016-18). A diagnosis of moderate-severe AS was based on standard clinical criteria: echocardiography (e.g., aortic valve area <1.5 cm², peak velocity ≥3.0 m/s, and mean gradient ≥30 mmHg), aortic valve replacement for severe AS or moderate AS in patients undergoing coronary artery bypass graft surgery, or a documented diagnosis of moderate or severe AS. Moderate-Severe AS was identified in a total of 142 participants.

Statistical Analysis

Participant demographic data was analyzed using descriptive statistics and stratified by AVC (AVC=0 vs. AVC>0) and age (age<65 years vs age ≥65 years). AVC score is 0 inflated and right skewed but normalizes with logarithmic transformation [ln (AVC+ 1)]. Therefore, for continuous analyses AVC was analyzed as a logarithmically transformed variable. AVC was also examined as a categorical variable (AVC 0, 1-99, ≥100 AU).

AVC progression was calculated as the change in AVC divided by the number of years between CT scans (2-10 years). Linear regression was used to examine the association between age (per 10 years older and age ≥65 years) and AVC progression. The age threshold of 65-years as the prevalence of AVC increases with age, particularly above the age range of 65 to 70 years old ^{22, 23}. We also conducted the analysis per 10 years older based on prior retrospective analyses showing two-fold increased risk of aortic valve disease per 10-year increase in age²⁴. An analysis for the association between age and AVC progression was also conducted after adjusting for baseline AVC. Linear regression models were adjusted for the following adjustment variables that were selected based on variables most likely to impact the association of AVC with AS: age, sex, field center, systolic blood pressure, diastolic blood pressure, hypertension medication, total cholesterol, HDL-C, lipid-lowering medication, fasting glucose, diabetes, body mass index, pack-years smoking, lipoprotein(a), education, and income¹⁴. Incident moderate-severe AS was selected as the primary outcome variable given strong association between AVC and long-term risk of severe AS. We also conducted linear regression analyses per one standard deviation higher annualized AVC progression. Cases of moderate AS were also included to maximize the power of the study and growing interest in earlier valve replacement strategies for patients with moderate AS, such as the ongoing PROGRESS trial²⁵ . Incident moderate-severe AS was assessed using event rate per 1,000 person-years and Kaplan-Meier survival curves, both stratified by age (age<65 years vs age≥65 years) and AVC categories. The association between AVC with incident moderate-severe AS was assessed using multivariable adjusted Cox proportional hazards ratios as follows: model 1: unadjusted, model 2: adjusted for age, race/ethnicity, systolic blood pressure, LDL-C, diabetes, lipoprotein(a) with fewer variables included compared to linear regression analyses to avoid overfitting given the relatively small number of events. We also performed sensitivity analyses to examine the progression of AVC 1) stratified by gender and 2) for participants with AVC=0 at baseline. We also examined the association of AVC with 1) moderate-severe AS excluding participants with bicuspid aortic valve and 2) with the outcome of all-cause mortality.

Results

Among both participants with AVC=0 and AVC>0, those who were age ≥65 years generally had a higher atherosclerotic burden (Table 1). However, older persons had lower low-density lipoprotein cholesterol and triglyceride levels and a higher proportion of lipid lowering medication.

Table 1.

Participant demographics stratified by age and absence or presence of AVC

	AVC = 0 (n=5,899)		p-value	AVC >0 (n=911)		p-value
	Age <65 (n=3,645)	Age ≥65 (n=2,257)		Age <65 (n=189)	Age ≥65 (n=722)
Age, years	54.3 (5.7)	71.4 (4.9)	<0.001	58.4 (4.3)	73.7 (5.3)
Women	54.3%	55.8%	0.261	32.8%	41.8%	0.024
Race
White	37.3%	37.6%	0.783	42.3%	46.1%	0.783
Black	27.7&	28.9%	0.287	24.4%	25.7%	0.287
Chinese	23.1%	20.2%	0.008	28.0%	20.3%	0.008
Hispanic	12.0%	13.3%	0.144	5.3%	7.9%	0.144
Systolic blood pressure, mm Hg	120.4 (18.7)	133.1 (22.3)	<0.001	130.1 (20.7)	136.3 (22.3)	<0.001
Diastolic blood pressure, mm Hg	72.4 (10.2)	71.0 (10.4)	<0.001	75.4 (10.4)	71.4 (9.8)	<0.001
Hypertension	46.7%	73%	<0.001	68.8%	80.9%	<0.001
Antihypertensive medication	23.4%	47.4%	<0.001	45.5%	58%	<0.001
LDL-C, mg/dL	118.0 (31.6)	115.3 (29.9)	0.001	129.2 (39.2)	115.9 (32.5)	<0.001
HDL-C, mg/dL	50.3 (14.6)	52.8 (15.2)	<0.001	47.0 (13.2)	49.5 (14.4)	0.036
Triglycerides, mg/dL*	112 (77, 164)	107 (76,155)	0.007	133 (92,186)	118 (82, 166)	0.017
Lipoprotein(a), mg/dL*	16.5 (7, 39)	17.5 (8,38)	0.296	25.4 (9, 67)	19 (8, 52)	0.027
Hyperlipidemia	43.5%	47.8%	<0.001	63.0%	53.6%	0.001
Lipid-lowering medication	11.0%	20.7%	0.001	22.0%	27.2%	<0.001
Glucose, mg/dL	95.3 (30.8)	98.6 (25.9)	<0.001	103.5 (40.4)	102.5 (35.7)	0.739
Diabetes	9.8%	14.1%	<0.001	21.2%	19.5%	<0.001
Body mass index, kg/m2	28.7 (5.7)	27.7 (5.2)	<0.001	30.0 (5.4)	28.1 (4.7)	<0.001
Current smoking	17.0%	7.6%	<0.001	20.1%	8.0%	<0.001
10-year ASCVD risk score, %	6.0 (6.1)	21.1 (12.9)	<0.001	12.0 (8.9)	28.0 (15.1)	<0.001
Coronary artery calcium *	0 (0, 9.6)	25 (0, 173)	<0.001	33 (0, 223)	155 (21, 536)	<0.001
Aortic valve calcium*	0	0	n/a	34.1 (13, 1,113)	69.0 (23, 2,453)	<0.001

^*Median and interquartile range

AVC progression

Participants <65 years and ≥65 years had no significant difference in median annualized AVC progression within the baseline AVC categories of 1-99 (10 versus 12 AU/year, p=0.846) and AVC>100 (50 versus 47 AU/year, p=0.303) (Central Illustration). We observed similar findings when stratified by sex (Supplemental Figure 1). In multivariable adjusted regression modeling, there was a greater annualized AVC progression per every 10 years of older age (β coefficient 1.43, p=0.006) and for participants ≥65 years compared to younger participants (β coefficient 26.37, p=0.001). (Table 2). However, after adjusting for baseline AVC, the associations were attenuated. The results remained significant within specific AVC groups, but the overall results showed no significant difference in AVC progression per 10-years older (β coefficient 0.32, p=0.511) or for older participants (β coefficient 9.67, p=0.134).

CENTRAL ILLUSTRATION:

Impact of age on aortic valve calcium progression and risk for aortic stenosis: Multi-Ethnic Study of Atherosclerosis

Table 2.

Linear regression (β-coefficient) of annualized AVC progression stratified by age and AVC group

	Overall	Baseline AVC 1-99	Baseline AVC ≥100
Per 10 years old
Unadjusted for baseline AVC	1.43 (p=0.006)	0.35 (p=0.002)	1.24 (p=0.474)
Adjusted for baseline AVC	0.32 (p=0.511)	0.31 (p=0.006)	0.44 (p=0.735)
Per 1 SD higher, adjusted for baseline AVC	0.003 (p=0.511)	0.04 (p=0.006)	0.01 (p=0.735)
Age ≥65 years old
Unadjusted for baseline AVC	26.37 (p=0.001)	5.51 (p=0.015)	49.65 (p=0.053)
Adjusted for baseline AVC	9.67 (p=0.134)	4.76 (p=0.034)	46.18 (p=0.038)
Per 1 SD higher, adjusted for baseline AVC	0.10 (p=0.134)	0.05 (p=0.034)	0.48 (p=0.038)

All models adjusted for: sex, race/ethnicity, field center, systolic blood pressure, diastolic blood pressure, hypertension medication, total cholesterol, HDL-C, LDL-C, lipid-lowering medication, lipoprotein(a), fasting glucose, diabetes, body mass index, pack-years smoking, education, and income.

In sensitivity analysis examining participants with AVC=0, 433 developed incident AVC >0. Among these participants, we found no significant difference in AVC progression based on age <65 versus ≥65 (β coefficient 8.97, p=0.079) or per every 10 years older age (β coefficient 0.716, p=0.176). We also found no significant difference in annualized AVC progression per every 10 years older age or for participants ≥65 years compared to younger participants for either women or men (Supplemental Table 1). There was no significant difference in AVC progression between older and younger participants, after exclusion of those with bicuspid valves (Supplemental Table 2).

Incident moderate-severe AS

Over a median follow-up of 16.7 years, 142 participants developed moderate-severe AS. Higher unadjusted event rates per 1,000 person-years for moderate-severe AS were observed among those age ≥65 years compared to those age <65 years within each AVC subgroup. However, the difference in moderate-severe AS event rate between older and young persons was only significant for those with AVC=0 and both age groups with AVC=0 had a very low event rate (Figure 1). In unadjusted Kaplan-Meier survival analysis, both participants age <65 years and age ≥65 years, had a significant reduction in survival free from moderate-severe AS for participants with AVC ≥100 compared to those with AVC=0. Participants aged ≥65 years had a slightly lower survival free from moderate-severe AS with AVC ≥100 at approximately 65% compared to approximately 75% among age <65 (Figure 2).

Figure 1.

Age specific incidence of moderate-severe AS by AVC categories per 1,000 person-years follow-up

Figure 2.

Kaplan-Meier curve: survival free from moderate-severe AS stratified by age and AVC category

Multivariable adjusted analysis demonstrated that AVC >0 was associated with a higher risk of long-term incident moderate-severe AS among younger HR 13.37 (95% CI 5.67-31.52) and older HR 10.59 (95% CI 6.77-16.56) participants. When analyzed as a categorical variable, higher AVC scores were significantly associated with an increased risk for moderate-severe AS among both younger and older participants, except for participants <65 years old with AVC 1-99 in whom there were only 2 moderate-severe AS events HR 2.68 (95% CI 0.54-12.52) (Figure 3).

Figure 3.

Association of aortic valve calcium with incident moderate-severe aortic stenosis stratified bv age

There was a significant association between AVC and all-cause mortality for older patients and for younger patients in unadjusted, but not adjusted results (Supplemental Table 3).

Discussion

In this study, we demonstrate similar progression of AVC for younger versus older persons and that the burden of AVC is the primary driver of AVC progression. Additionally, our results show that AVC is predictive of incident long-term moderate severe AS for both younger and older persons. Accordingly, these results provide insight into the similar underlying pathophysiology for CAVD in both younger and older persons and further highlights the prognostic utility of AVC to identify persons at increased risk for moderate-severe AS regardless of age.

The prognostic role of CAVD severity has been reinforced using echocardiographic data as baseline hemodynamic severity of AS has been shown to predict faster progression to severe AS ^26–29. Older age has also been associated with incident AS using echocardiographic data. However, given age’s association with a higher burden of CAVD²³ and the lack of adjustment for baseline CAVD disease severity in these studies, it is likely that the baseline severe CAVD of the older sample accounts for the observed association of age with incident AS ^30–32. Echocardiography has demonstrated that greater baseline AVC predicts faster progression to severe AS and a higher mortality risk among patients with mild or moderate disease²⁸. Our study, using CT measured AVC, further supports this finding by demonstrating that AVC burden is the predominant factor associated with rate of AVC progression regardless of age.

Our findings of no difference in AVC progression between age groups after adjusting for baseline AVC are consistent with other studies that have reported baseline AVC to predict AVC progression, not age ²² and extends these prior reports of 5-year progression data to 10-years. There are some studies that do report an association between age and disease progression ^{30, 33}, but these studies used echocardiogram evaluation of patients with AS, which has several potential limitations, particularly among patients with low flow, low gradient phenotypes, which can make the diagnosis of AS challenging.

In contrast, CT-quantified AVC is not subject to discordant hemodynamics and the use of sex-specific AVC thresholds (1,200 AU in women and 2,000 AU in men) has been recommended as a tie breaker for the diagnosis of severe AS ^{34, 35}. Additionally, our study sample included a wider age range of 45 to 84 years and adjusted for cardiovascular risk factors and baseline AVC. Our findings replicate a similar phenomenon observed for coronary heart disease where CAC scores have been found to be the predominant predictor of CAC progression beyond age^36–39.

We found a significant association between AVC and all-cause mortality for older patients, but only in unadjusted results for younger patients. However, the hazard ratio point estimates were similar for younger and older participants in the unadjusted results. The lack of a significant association among younger participants is likely due to lower power as among participants with AVC >0 there were only 37 deaths during follow-up for younger participants compared to 379 for older participants. These results are consistent other studies that have similarly shown a significant association of AVC with all-cause mortality⁴⁰.

In addition to age, several prospective analyses have examined the impact of sex on disease progression. The COFRASA-GENERAC study found faster disease progression in females compared to males after adjusting for baseline AVC⁴¹. In contrast, in the PROGRESSA study, AVC progression was reported to be 2-fold slower among females compared to males⁴². The discordance in findings between these studies and our study which showed no difference in AVC progression between sexes can be explained by several differences in the participants and methodology.^{26, 43, 44}. The participants included in the COFRASA-GENERAC study had higher baseline median AVC scores (median AVC 1,168 AU) compared to our study, and even included participants with severe AS. The faster progression reported may be secondary to the higher baseline AVC scores. The participants of the PROGRESSA study had mild AS or low-gradient AS and the researchers examined AVC progression thresholds to identify moderate hemodynamic progression, which poses diagnostic limitations among those patients with low-gradient AS. Additionally, our study had a longer follow-up period of a median of 16.7 years, while the follow up period in the COFRASA-GENERAC study was 3.2 years and in the PROGRESSA study was 2 years.

While there are currently no approved therapies for slowing AVC or preventing AS, our study provides AVC progression reference ranges that may be clinically helpful for estimating disease progression and timing of follow-up surveillance imaging. This framework may be useful for clinicians particularly when evaluating patients who are earlier in the disease process and asymptomatic. Our data suggests that AVC 1-99 is associated with approximately 10 AU/year AVC progression, while AVC >100 is associated with approximately 50 AU/year AVC progression. Therefore, it may be reasonable to consider follow-up echocardiogram imaging in 5-10 years among those with AVC 1-99, and 3-5 years among those with AVC >100 or even shorter for those with very high AVC scores. Combining our findings with prospective analyses on the impact of sex on AVC progression, it may be reasonable to consider even more frequent surveillance if the patient is female. Additionally, given the up to 75-fold increased risk of AS, the presence of AVC >100 may warrant consideration of echocardiography if not previously performed.

Study Limitations

Limitations of this study include that not all participants had a follow-up CT scan and the relatively small number of participants who developed moderate-severe AS, particularly for those age <65 years old, which reduces the precision of our event rate, hazard ratio estimates, and confidence intervals. However, the lower limits of our 95% CI still demonstrate a significant association between AVC and moderate-severe AS. We acknowledge that use of another age cutoff could be justified. The small sample size also limits our ability to conduct further subgroup analysis, such as finer delineation of age categories. Lastly, our Cox model did not adjust for bicuspid aortic valves, however, sensitivity analysis found similar results after excluding patients with bicuspid AS (n=10) (Supplemental Table 2). Strengths of the study include the use of non-contrast CT for the measurement of AVC and the adjudication of severe AS⁴⁵, as well as the long follow up period for AVC progression analyses of 10 years and severe AS analyses of 16.7 years.

Conclusion

Our study found a similar rate of progression of AVC among younger and older age persons, which was predominantly based on baseline AVC burden. Additionally, our results demonstrate that AVC >0 is associated with a higher risk of long-term incident moderate-severe AS among both older and younger populations. These findings emphasize the prognostic importance of AVC in moderate-severe AS among individuals of all ages. Therefore, CT measured AVC is an important tool that should be more frequently utilized for risk stratification and considered for use in patient selection for novel clinical trials and therapeutics aiming to slow down disease progression and prevent moderate-severe AS.