Incremental Value of Aortomitral Continuity Calcification for Risk Assessment after Transcatheter Aortic Valve Replacement

Study Population

The Stanford Cardiovascular Institute institutional review board approved this Health Insurance Portability and Accountability Act–compliant retrospective study. The requirement to obtain written informed consent was waived. All patients who underwent TAVR between March 2013 and March 2016 were eligible. Only patients without prior valve replacement who underwent preprocedural unenhanced and contrast material–enhanced CT of the aorta were included. Preprocedural demographic and clinical information, including cardiac risk factors, STS score, and frailty index (based on gait speed, Katz activities of daily living score, serum albumin level, and grip strength), as well as procedural details and clinical follow-up data were collected from electronic medical records.

CT Examination

All patients underwent electrocardiography-synchronized unenhanced and contrast-enhanced aortic CT. Tube voltage and current varied according to patient size. Thin-section images were selected for evaluation. The calcification burden of the aortomitral continuity as a whole (AMCC mass, including the aortic valve, left ventricular outflow tract, and mitral annulus and/or valve), isolated aortic valve (AVC mass), and coronary arteries (Agatston score and CAC score) were quantified from unenhanced CT scans by using commercially available software (Aquarius iNtuition; TeraRecon, Foster City, Calif) (Fig 1, Fig E1 [supplement]). Because calcium attenuation is influenced by tube voltage, we used a calcium threshold of 130 HU for tube voltages of at least 120 kVp. For tube voltages less than 120 kVp, we used a calcium threshold of 147 HU (8). Previous studies have shown that the quantification of aortic and mitral valve calcification with three-dimensional workstations is fast and highly reproducible (6,9). To correct valve calcifications for aortic annulus size, we normalized AMCC and AVC for the aortic annulus area, measured on the contrast-enhanced CT scans. CT measurements were performed separately by two observers with more than 7 years of experience in cardiovascular imaging (M.J.W. and E.M., 8 and 20 years of experience, respectively). Observers were blinded to clinical, echocardiographic, and outcome data.

Figure 1a:

Download as PowerPoint

Figure 1b:

Download as PowerPoint

Echocardiography

Patients underwent transthoracic echocardiography before and after the procedure before hospital discharge. Echocardiographic evaluation was performed according to the European Association of Echocardiography guidelines (10). Mean aortic valve gradient, aortic valve area, peak aortic jet velocity, left ventricular ejection fraction, and right ventricular systolic pressure were assessed. Postprocedural echocardiograms were reviewed by an independent experienced observer for the presence and severity of PVL and were classified as none, mild, moderate, or severe (11).

Clinical Outcomes

The primary end point was defined as all-cause mortality 1 year after TAVR. Secondary end points were prolonged hospital stay (hospitalization >7 days) and postprocedural paravalvular aortic regurgitation (mild-to-severe PVL as determined with postprocedural echocardiography) (11).

Statistical Analysis

Data are presented as medians (first quartile to third quartile) for continuous variables and as numbers of patients and percentages for categorical variables. We compared baseline data between survivors and nonsurvivors by using the Mann-Whitney U test for continuous data and the Pearson χ² test for categorical data when no cell counts less than five present, the Fisher exact test for categorical data when cell counts less than five were present, or the Kendall tau-c for ordinal data.

We tested associations between baseline variables and 1-year all-cause mortality with univariable Cox regression analyses. For multivariable analyses, we evaluated the added prognostic value of the CAC score, AVC mass, and AMCC mass beyond traditional clinical factors. In the multivariable models, we only used the STS risk score (continuous variable) and not dichotomous STS score of at least 8, owing to potential multicollinearity considerations. The multivariable model was corrected for overoptimism by using internal validation with bootstrap analysis at 1000 iterations adjusting for age, sex, cardiovascular risk factors (hypertension, diabetes, hyperlipidemia, and smoking status), frailty index, New York Heart Association classification, and valve type. On the basis of the prediction models, the hazard ratios of the predictors were calculated with 95% confidence intervals. We used the Youden index to determine the optimal thresholds for Kaplan-Meier survival analyses. Survival curves were compared by using the log-rank test. To evaluate the discriminative capacity of the models to predict mortality, we calculated concordance statistics and net reclassification improvement indexes for four risk categories: less than 6%, 6%–10%, 10%–20%, and greater than 20% (12,13). Area under receiver operating characteristic curve values were compared by using the method described by DeLong et al (14).

To evaluate the secondary end points, we used univariable logistic regression analyses, with postprocedural aortic PVL and prolonged hospital stay as outcome variables. For multivariable analyses, we evaluated the added prognostic value of the CAC score, AVC mass, and AMCC mass beyond traditional clinical factors. The multivariable models were adjusted for age, sex, cardiovascular risk factors (hypertension, diabetes, hyperlipidemia, and smoking status), frailty index, the New York Heart Association classification, and valve type. Odds ratios of the predictors were calculated with 95% confidence intervals.

P < .05 was considered indicative of a statistically significant difference. Statistical analyses were performed with R version 3.3.3 (R Foundation for Statistical Computing, Vienna, Austria) with the survival, rms, and nricens packages, SPSS version 24 (IBM SPSS Statistics, IBM, Armonk, NY), and MedCalc version 17.9.7 (MedCalc, Ostend, Belgium).

All-Cause Mortality

Results of primary outcome analyses are provided in Table 4. STS score, CAC score, and AMCC mass were the strongest predictors for all-cause mortality within 1 year after TAVR at univariable Cox regression analysis. The same variables remained associated with all-cause mortality within 1 year after TAVR in the adjusted multivariable model corrected for overoptimism. AVC mass was not predictive of mortality within the 1st year after TAVR. AMCC mass and CAC score did not add discriminative power over the basic model, despite having slightly higher C statistic values (Table 5). However, adding these variables together did result in improved reclassification of patients, with net reclassification improvement indexes of 14%–25%.

Table 4: Cox Proportional Hazards Regression Analysis with All-Cause Mortality within 1 Year as the Outcome Variable (n = 329)

Table 5: Concordance Statistics and Net Reclassification Improvement Analysis with All-Cause Mortality within 1 Year as Outcome Variables (n = 329)

Optimal thresholds for a CAC score of 2001, AVC mass of 300 mg, and AMCC mass of 700 mg derived from receiver operating characteristic curve analysis were used in the Kaplan-Meier survival analyses (Figs 2, 3). One-year survival was significantly worse with AMCC mass greater than 700 mg (P < .001), AVC mass greater than 300 mg (P = .0496), CAC score greater than 2001 (P = .0019), and STS score of at least 8 (P = .0018). After 2 years, the differentiation for AMCC mass became more prominent compared with other variables (Fig E2 [supplement]). One-year mortality in high-risk patients with an STS score of at least 8 was 22% (28 of 130 patients), which increased to 28% (19 of 67 patients) when these patients had an AMCC mass greater than 700 mg. One-year mortality in low-risk patients with an STS score lower than 8 was 9% (18 of 199 patients), which decreased to 6% (seven of 117 patients) with an AMCC mass of 700 mg or less.

Figure 2:

Download as PowerPoint

Figure 3:

Download as PowerPoint

Prolonged Hospital Stay and Postprocedural PVL

Results of the secondary outcome analyses are shown in Table 6. Postprocedural echocardiography could not be performed in three patients who died shortly after TAVR. Mild-to-severe aortic PVLs were found in 147 of the 329 patients (45%) and moderate-to-severe PVLs were found in 32 (10%). Age, STS score, baseline mean aortic valve area, and baseline AVC mass were significantly associated with aortic PVLs at univariable logistic regression analysis. AVC mass was associated with postprocedural aortic PVLs (odds ratio: 1.93 [95% confidence interval: 1.39, 2.48] per 100 mg, P = .001). AMCC mass and CAC score were not significantly associated with PVLs; however, AMCC mass and AVC mass were higher at increasing aortic regurgitation (Table E1 [supplement]).

Table 6: Logistic Regression with Postprocedural Aortic Regurgitation and Prolonged Hospital Stay as Outcome Variables

Fifty-two of the 329 patients (16%) were hospitalized for more than 7 days after TAVR. Logistic regression showed that only the STS score was associated with prolonged hospital stay (odds ratio: 1.19 [95% confidence interval: 1.08, 1.32], P < .001).

Cardiovascular Calcifications

Evaluation of the prognostic value of cardiovascular calcifications, such as AVC and mitral annulus calcification, has recently generated substantial interest in patients with aortic stenosis. AVC has been shown to be associated with postprocedural PVLs, procedural complications, and 30-day, 1-year, and long-term mortality (5,15,16). We also found a significant association of AVC with PVLs, which can be explained by its location in the landing zone of the device with direct consequences on the deployment, expansion, and apposition of a stent-mounted valve. However, we did not find a significant association between AVC and mortality in the 1st year after TAVR. Only our 2-year survival analysis (Fig E2 [supplement]) suggests that AVC mass is a better predictor of long-term mortality than of short-term mortality. Many studies have evaluated the prognostic value of mitral annulus calcification in patients undergoing TAVR and have shown associations with increased all-cause and cardiovascular mortality (6,17).

We used AMCC as an aggregate quantitative measure of cardiac and/or valvular calcifications, including valve leaflets and/or annulus, and interspersed tissue between the aortic and mitral valves. AMCC is easy to measure and does not require time-consuming separation of a calcified valve and/or annulus from adjacent fibrous tissue calcifications—a common occurrence in patients with severe aortic stenosis. Correcting for body size is straightforward. We used the readily available aortic valve area, which is routinely measured in candidates for TAVR. Although leaflet calcifications are important in patients with aortic stenosis, we hypothesized that the total burden of calcifications might be a stronger marker of cardiovascular risk and mortality because of a presumed common pathway leading to active cardiovascular mineral deposition. In contradistinction to AVC, AMCC was a strong, significant predictor of mortality 1 year after TAVR in our cohort. The association between AMCC and 2-year mortality may be even stronger (Fig E2 [supplement]).

Another parameter that turned out to be a significant predictor of mortality the 1st year after TAVR was CAC. When adding CAC greater than 2001 and AMCC greater than 700 mg to high-risk patients (STS score >8), the 1-year mortality increased from 22% to 34%. If confirmed in other cohorts, such an estimate may help recalibrate patient and family expectations and guide the difficult process of decision-making in these vulnerable patients.

The different associations of AMCC and/or CAC and AVC indicate that these calcifications may be either driven by disparate pathways or are merely a reflection of their respective anatomic locations.

Limitations

Our findings are based on a single-center retrospective cohort of patients undergoing TAVR, with 40% of the individuals at high surgical risk (STS score >8). Although the distributions of STS scores will obviously be different in lower-risk populations, the observed associations between CT-detected calcifications and mortality may still be preserved and allow identification of those individuals who benefit least from TAVR. The secondary outcome, prolonged hospital stay, included a heterogeneous group of patients with only a prolonged hospital stay and patients who died within 7 days. The latter only concerned three patients in our study, and repeat analyses after excluding these patients did not substantially change results. Another limitation was that variables such as preprocedural diffusing capacity of the lungs for carbon monoxide and right ventricular systolic pressure were not obtained in all patients. Despite the finding that these differed significantly between survivors and nonsurvivors, the reduced sample size prevented the use of these variables in survival prediction modeling.

Conclusion

AMCC is a predictor of mortality within 1 year after TAVR and, together with CAC, substantially improves individual risk classification when added to the clinically used STS score.