Diastolic Filling Parameters Derived from Myocardial Perfusion Imaging Can Predict Left Ventricular End-Diastolic Pressure at Subsequent Cardiac Catheterization

MATERIALS AND METHODS

MPI

All patients underwent dual-isotope MPI with 111–148 MBq of ²⁰¹Tl at rest and 925–1,110 MBq of ^99mTc for stress acquisition. Fifteen minutes after ²⁰¹Tl-labeled tracer injection, resting SPECT images were acquired. Sixty minutes after ^99mTc-labeled tracer injection, at peak stress, electrocardiogram-gated SPECT images were obtained. A fixed 90° dual-detector camera (C-cam; Siemens Medical Systems) with elliptic 180° arc (45° right anterior oblique to 45° left posterior oblique), 32 projections at 30 s per projection, and 16 frames per R–R interval was used to acquire SPECT data. The projection data were reconstructed into tomographic transaxial images by use of a filtered backprojection technique with a 3-dimensional ramp Butterworth filter (order, 4; cutoff frequency, 0.50 cycle per pixel). For data analysis, 4D-MSPECT software (University of Michigan) was applied to determine the left ventricular ejection fraction (LVEF), end-diastolic volume (EDV), and end-systolic volume. 4D-MSPECT software automatically fits the left ventricular volume curve (Fig. 1) with a fourth-order harmonic function and derives peak filling rate (PFR), time to PFR (TPFR), and filling rate during the first third of diastole (1/3FR). All diastolic measurements were obtained during stress-gated ^99mTc MPI.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

Left ventricular diastolic filling curves derived from 4D-MSPECT. (A) Curve with normal diastolic filling parameters. (B) Curve with abnormal diastolic filling parameters. LVEDPs determined at subsequent cardiac catheterization are shown. bpm = beats per minute; EF = ejection fraction; ESV = end-systolic volume; HR = heart rate; LV = left ventricular; MFR3 = filling rate during first third of diastole; PER = peak emptying rate; TPER = time to peak emptying rate; vol = volume.

RESULTS

4D-MSPECT Diastolic Function Parameters and LVEDP

The association between diastolic function parameters from 16-frame gated SPECT and LVEDP is shown in Figure 2. There were significant inverse correlations between LVEDP and PFR (r = −0.53, P = 0.00005) and between LVEDP and 1/3 FR (r = −0.45, P = 0.0009). There was also a significant positive correlation between LVEDP and TPFR (r = 0.45, P = 0.0009). Correlations remained significant and similar in the presence or absence of CAD (r = 0.42–0.57, P < 0.05), except that 1/3FR did not achieve statistical significance in patients without CAD.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

Correlations between 4D-MSPECT diastolic parameters and LVEDP.

Resting ²⁰¹Tl LHR and LVEDP

The correlation between LVEDP and resting ²⁰¹Tl LHR was not statistically significant (r = 0.10, P = 0.49) (Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3. 

Correlation between LVEDP and resting ²⁰¹Tl LHR.

ROC Curve Analysis

The ROC curves for detecting LVEDPs of ≥18 mm Hg or <18 mm Hg with PFR, TPFR, and 1/3FR are shown in Figure 4. The areas under the curve (AUCs) for PFR, TPFR, and 1/3FR were 0.83, 0.75, and 0.80, respectively. The optimal PFR, TPFR, and 1/3FR cutoff values for providing maximum sensitivity and specificity were ≤2.57 end-diastolic volume per second (EDV/s), ≥161 ms, and ≤1.52 EDV/s, respectively.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4. 

ROC curve analysis for ability of PFR (A), TPFR (B), and 1/3FR (C) to detect LVEDPs of ≥18 mm Hg or <18 mm Hg.

These PFR, TPFR, and 1/3FR values showed sensitivities of 82%, 88%, and 82% and specificities of 80%, 63%, and 74%, respectively, for correctly identifying patients with LVEDPs of ≥18 mm Hg. Each diastolic filling variable showed a positive predictive value (PPV) of ≥54% and a negative predictive value (NPV) of ≥90%.

Hence, all diastolic filling parameters improved the ability to detect LVEDPs of ≥18 mm Hg (AUCs of ≥0.75). These parameters showed good sensitivity (≥82%), but the specificity (≥63%) and the PPV (≥54%) were modest. The NPV of these parameters for identifying patients with LVEDPs of <18 mm Hg was remarkably good (≥90%).

After identifying these 3 variables as predictors of diastolic dysfunction, we investigated the incremental effect of the presence of more than one variable. TPFR correlated well with PFR and 1/3FR (r = −0.32 and −0.64, respectively; P = 0.02 and <0.0001, respectively), whereas PFR did not correlate with 1/3FR (r = 0.18, P = 0.19). Because there was no correlation between PFR and 1/3FR, it was feasible to use these 2 parameters in combination to assess their combined predictive power for LVEDPs of ≥18 mm Hg. When the results for both were above the cutoff values, PFR and 1/3FR combined showed a PPV of 86%, a sensitivity of 67%, and a specificity of 94% for detecting LVEDPs of ≥18 mm Hg. Similarly, the normal values for both of these parameters showed an NPV of 84% for detecting LVEDPs of <18 mm Hg. All values from the ROC curve analysis are shown in Table 2.

View this table:

• View inline
• View popup

TABLE 2

ROC Curve Analysis, Sensitivity, Specificity, PPV, and NPV

DISCUSSION

The predictive power of MPI for subsequent cardiac events and the severity of CAD is significantly enhanced when stress ²⁰¹Tl lung uptake is added to the extent and severity of perfusion defects (4,18–20). Several studies have shown that increased stress ²⁰¹Tl lung uptake correlates very well with increased left ventricular filling pressure (4–7). Elevated LVEDP is a major predictor of left ventricular diastolic dysfunction (15,16) and has been shown to be an independent predictor of future clinical heart failure events (21) and mortality (1,22). In these circumstances, the ability to predict an elevation of LVEDP during MPI could improve the predictive power of myocardial perfusion studies for subsequent cardiac events.

At the present time, the majority of MPI studies are performed with stress ^99mTc-labeled perfusion agents. With the transition from ²⁰¹Tl to ^99mTc, access to stress ²⁰¹Tl lung uptake data is no longer available. With the loss of stress ²⁰¹Tl LHR data, few studies have assessed the utility of stress ^99mTc and resting ²⁰¹Tl lung uptake for predicting left ventricular filling pressure (8–10,23).

Studies with ^99mTc MPI have shown a very poor correlation between stress ^99mTc LHR and elevated left ventricular filling pressure (9,10), probably because of the rapid washout of ^99mTc from the lung during stress (24). Marcassa et al. (23) investigated the use of resting ²⁰¹Tl LHR, which is quite often used with ^99mTc in dual-isotope imaging. In a group of 159 consecutive patients with LVEFs of ≤35% (mean ± 1 SD, 23% ± 6%), resting ²⁰¹Tl LHR had significant predictive power for subsequent cardiac events and mortality. Resting ²⁰¹Tl LHR also correlated well with pulmonary capillary wedge pressure in 19 patients. This approach appeared to show some promise, especially because Patel et al. (10) showed a significant correlation between resting ²⁰¹Tl LHR and stress ^99mTc LHR (r = 0.55, P < 0.01). However, correlation coefficients for stress ^99mTc LHR with LVEDP and with pulmonary capillary wedge pressure were very modest, at 0.29 and 0.34, respectively. Despite the strong correlation between resting ²⁰¹Tl LHR and stress ^99mTc LHR, multivariate linear regression indicated that the relationship between pulmonary capillary wedge pressure and LVEDP was significant with stress ^99mTc LHR but not with resting ²⁰¹Tl LHR. Given the modest correlations and the rapid washout of ^99mTc, the reporting of LHR with stress ^99mTc is not a common practice. LHR with resting ²⁰¹Tl appears to be somewhat inferior to LHR with stress ^99mTc but, more importantly, it has never been validated in a group of patients with preserved left ventricular systolic function (8,10,23).

With the improved count statistics and resolution of MPI with a ^99mTc-labeled radiotracer, LVEF and wall motion can be reliably assessed during MPI (25,26). More recently, diastolic filling curves were derived from an assessment of wall motion during diastole (13,14,27,28) with 16-frame acquisitions. Studies with gated MPI have shown that PFR, TPFR, and 1/3FR can be derived and show excellent correlations with the same markers derived from radionuclide angiography (13,14). Furthermore, these diastolic function parameters have been shown to correlate well with Doppler echocardiographic variables of diastolic dysfunction (28) and with plasma B-type natriuretic peptide levels (28). 4D-MSPECT software derives the second-derivative curve from wall motion and automatically calculates PFR, TPFR, and 1/3FR. We are not aware of any studies in which 4D-MSPECT was used to assess left ventricular diastolic parameters at MPI and correlating these parameters with LVEDP at subsequent cardiac catheterization.

Three diastolic parameters measured by 4D-MSPECT showed a robust correlation with LVEDP at subsequent cardiac catheterization within 15 d and without an interval event, an intervention, or a change in blood pressure medication. PFR and 1/3FR correlated negatively with LVEDP. An inverse correlation of decreasing values for PFR and 1/3FR with increasing values for LVEDP is expected in diastolic dysfunction, in which the filling rate decreases as LVEDP increases. Conversely, as diastolic dysfunction worsens, TPFR increases, as reflected by the positive correlation between increasing values for TPFR and LVEDP. Interestingly, resting ²⁰¹Tl LHR was not correlated with LVEDP when LVEFs were ≥45%. These observations suggest the superiority of 4D-MSPECT diastolic filling variables over resting ²⁰¹Tl lung uptake for assessing diastolic function.

We also evaluated the diagnostic utility of PFR, TPFR, and 1/3FR for detecting specific levels of LVEDP in patients with preserved systolic function. ROC curve analysis showed that the AUC was ≥0.75 for each of these parameters for detecting LVEDPs of ≥18 mm Hg. Individual parameters showed excellent sensitivity (≥82%) and modest specificity (≥63%), whereas the combination of abnormal values for PFR and 1/3FR showed a specificity of 94% for determining LVEDPs of ≥18 mm Hg. Hence, these parameters will provide excellent screening individually and, when combined, will represent a powerful diagnostic tool for the assessment of diastolic function during MPI.

The era of planar imaging and stress–redistribution ²⁰¹Tl protocols enabled us to obtain stress ²⁰¹Tl LHR, which provided significant predictive power for subsequent cardiac events. Our findings indicate that diastolic filling variables could improve the predictive power of MPI by allowing the integration of data on diastolic impairments with data on the severity and extent of myocardial perfusion defects. This integration could be applied to the new era of stress ^99mTc-labeled agents and SPECT MPI. This investigation was performed with equipment that is either present or easily available in every clinical nuclear cardiology laboratory performing SPECT MPI. We believe that values of PFR, TPFR, and 1/3FR should be reported for all SPECT MPI studies to allow the detection of diastolic dysfunction in the presence or absence of CAD.

The novel aspect of the present study is that it is directly and widely applicable to the general population of patients being evaluated by MPI for suspected CAD. The vast majority of these patients have preserved systolic function and sometimes present with symptoms of dyspnea, which is considered an equivalent of angina (29,30). Patients with preserved systolic function generally have lesser degrees of diastolic dysfunction, unlike patients who were evaluated in previous nuclear studies (8,10,23) of LHR. We believe that our study is the first to combine a methodology for assessing myocardial perfusion with a methodology for detecting diastolic dysfunction in patients with preserved LVEF. If our retrospective findings are confirmed prospectively, then this technique can be applied immediately to most patients being evaluated for the presence of CAD in nuclear cardiology laboratories.

The use of left ventricular filling curve parameters to assess diastolic function is not helpful for patients with irregular heart rhythm because gating is disabled. Our patient population underwent MPI and cardiac catheterization within 15 d. This delay could have affected the volume status of the patients and therefore the LVEDPs measured at different time points. Despite this limitation, we found good correlations between diastolic function indices from MPI and LVEDPs. Because our study was retrospective, we relied on the accuracy of medical records. However, tracings of all LVEDP readings were manually checked, and nuclear studies were reprocessed and reanalyzed. Because the present study was a retrospective series of consecutive patients, a prospective investigation of this technique is indicated to validate its utility in predicting LVEDP at cardiac catheterization.

CONCLUSION

In patients with LVEFs of ≥45%, resting ²⁰¹Tl LHR did not correlate significantly with LVEDP. Diastolic function parameters (PFR, TPFR, and 1/3FR) derived from 16-frame gated SPECT with 4D-MSPECT correlated significantly with LVEDP at cardiac catheterization. If the assessment of diastolic function is combined with MPI results for risk stratification, then 4D-MSPECT diastolic filling parameters are superior to resting ²⁰¹Tl LHR, according to our observations. These parameters showed good sensitivity and specificity for detecting LVEDPs of ≥18 mm Hg and a good NPV for identifying patients with LVEDPs of <18 mm Hg. Combining PFR and 1/3FR can result in improved PPV (86%) and specificity (94%).