Prognostic Value of Lung Sestamibi Uptake in Myocardial Perfusion Imaging of Patients With Known or Suspected Coronary Artery Disease

Study population

Between January 1994 and April 1999, consecutive patients who underwent stress and rest ^99mTc-sestamibi single-photon emission computed tomographic (SPECT) myocardial perfusion imaging on the same imaging system (Elscint 409; Haifa, Israel) were considered for inclusion in the study cohort (n = 1,027). The study cohort, stress procedures, and image processing used in our laboratory have been previously described (6). Exclusion criteria included planar imaging or imaging not completed on the designated camera (n = 36), corrupted (n = 102) or poor quality image data (n = 2), nonstandard stress procedures (n = 4), or the inability to link patient data to the Manitoba Population Health Research Data Repository (n = 140). For patients undergoing more than one scan during the study period, only the first scan was included in the analysis (n = 25). The study protocol was approved by the local Research Ethics Board and provincial Health Information Privacy Committee.

Stress protocols and imaging

A two-day protocol utilizing treadmill exercise is the preferred procedure in our laboratory. Whenever possible, beta-blockers and calcium channel antagonists are withheld for 24 to 48 h before the stress procedure, and nitrates are avoided for at least 6 h. Symptom-limited treadmill exercise was performed with tracer injection at peak exercise. Individuals unable to achieve a satisfactory exercise workload underwent pharmacologic stress with dipyridamole. Supine, nongated SPECT imaging commenced 30 to 60 min after stress and 45 to 75 min after the resting injection.

Scintigraphic interpretation

Initial visual interpretation of the scan data was performed without image quantification using two variables: a normal scan (normal or equivocal) versus abnormal scan (abnormal with fixed or reversible defects); and no reversibility (normal, equivocal, or abnormal with fixed defects) versus reversibility (abnormal with fully reversible or partially reversible defects). Image data were subsequently reprocessed using commercial software (AutoSPECT and QPS AutoQUANT, Cedar Sinai Medical Center and ADAC Laboratories, Milpitas, California). Left ventricular contours were checked visually and manually adjusted if the computer-generated automatic contours were found to be incorrect. This software provides quantitative defect scores for the summed stress score (SSS), summed rest score (SRS), and summed difference score (SDS) based on a 20-segment model (7,8). We have previously shown close agreement between the visual and automated quantitative assessment in terms of diagnosis and prognosis (6,9). Visual interpretation and the quantitative analysis were performed without knowledge of patient outcomes.

Lung uptake of ^99mTc-sestamibi was calculated as the lung-to-heart ratio (LHR), using a validated automated technique (3). The LHR was measured on stress (sLHR) and rest projection data (rLHR), and the difference (dLHR) was derived by subtraction. Mean sLHR was identical for exercise only and pharmacologic stress, and these groups were combined in subsequent analysis. Excellent interobserver agreement in LHR measurements was confirmed in 30 blindly reprocessed cases (sLHR: R = 0.97, p < 0.00001, mean difference 0.01; and rLHR: R = 0.95, p < 0.00001, mean difference 0.01).

An automated calculation of ungated LV volumes after stress and at rest was obtained (10). Transient ischemic dilation was derived from the endocardial volumes as the ratio of LV volume at stress divided by the volume at rest.

Outcomes assessment

Cohort follow-up was performed through the Manitoba Population Health Research Data Repository, which links computerized data bases of physician services and hospitalizations for all residents of the province of Manitoba (11). The primary outcome was death (from Vital Statistics) or acute myocardial infarction (AMI; hospital diagnosis ICD-9-CM 410.xx). Clinical and stress variables were primarily obtained from chart review. The Repository was used to identify diabetes or previous AMI.

Statistical analysis

Statistical analysis was performed with a commercial software package (Statistica Version 6.1, StatSoft Inc., Tulsa, Oklahoma). Continuous variables are reported as the mean value ± SD, and p < 0.05 is considered to represent a statistically significant difference. Group comparisons were performed using analysis of variance (ANOVA) or chi-square testing. Kaplan-Meier survival curves were compared using Gehan℉s Wilcoxon test. Receiver operating characteristic (ROC) curves were generated with SPSS version 11.0 (SPSS Inc., Chicago, Illinois). Multivariate survival analysis used Cox proportional hazards modeling. One model was developed for the complete study cohort using clinical variables, stress modality, perfusion, LV volume, and LHR measurements. A second model was developed for those undergoing treadmill exercise only and also included exertional chest pain, electrocardiographic ischemia, peak workload, peak heart rate, peak systolic blood pressure, and peak double product. An initial (partial) regression was limited to competing variables from within each of the clinical, exercise, volume, and lung uptake variable subgroups in order to select the final (combined) regression variables. If significant collinearity was identified (R > 0.5), only the factor with the strongest association was entered.

Patient characteristics

Of the 718 patients included in the analysis, 380 (53%) were males, 186 (26%) had a history of MI, 112 (16%) had undergone a previous coronary artery bypass graft surgery or coronary angioplasty, 90 (13%) were diabetic, and 556 (77%) had treadmill exercise as the only stress method.

Lung uptake measurements

Overall, mean sLHR (0.32 ± 0.06) was slightly greater than the rLHR (0.31 ± 0.06; p = 0.001 by paired comparison). Lung uptake measurements showed significant differences according to the severity of perfusion abnormality, as reflected in the SSS, with greater stress and rLHR in association with more extensive perfusion abnormalities (Table 1).

Survival analysis

During a mean follow-up of 5.6 ± 1.1 years, 114 patients (16%) experienced a primary event. There were 81 deaths and 62 AMIs, including 29 patients recorded as having AMI followed by death.

Mean sLHR was higher in individuals who had an adverse outcome, as compared with those who remained event-free (0.36 ± 0.08 vs. 0.31 ± 0.06, p < 0.00001). This was even identified in the subgroup with normal SSS (0.33 ± 0.06 vs. 0.30 ± 0.05, p = 0.002). Two-way ANOVA confirmed higher sLHR in relation to an adverse clinical outcome (p < 0.00001), independent of the SSS category. Individuals with an adverse outcome also had a higher mean rLHR (0.34 ± 0.07 vs. 0.31 ± 0.06, p < 0.00001) and dLHR (0.016 ± 0.057 vs. 0.004 ± 0.050, p = 0.02).

The ROC analysis confirmed a significant association between adverse outcomes over the range of sLHR measurements (Fig. 1).The area under the curve was 0.65 for sLHR (95% confidence interval [CI] 0.59 to 0.71, p < 0.0001), 0.62 for rLHR (95% CI 0.57 to 0.69, p < 0.0001), and 0.56 for dLHR (95% CI 0.51 to 0.62, p = 0.04). For comparison, the area under the curve was 0.67 for SSS (95% CI 0.61 to 0.73), 0.67 for SRS (95% CI 0.61 to 0.73), and 0.63 for SDS (95% CI 0.57 to 0.68). The area under the curve for sLHR was similar for the individual end points of death (0.65, 95% CI 0.58 to 0.72, p < 0.001) and AMI (0.67, 95% CI 0.60 to 0.74, p < 0.001). Subgroup analysis for individuals with a normal SSS gave an area under the curve for sLHR (0.64, 95% CI 0.52 to 0.75, p = 0.01), similar to the overall group. Event-free survivals according to sLHR tertiles (Fig. 2)were significantly different (p < 0.00001).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Receiver operating characteristic analysis for predicting death or acute myocardial infarction from stress lung-to-heart ratio (LHR) (solid line)or summed stress score (SSS) (dotted line).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Kaplan-Meier plots of event-free survival according stress lung-to-heart ratio (LHR) tertile.

Table 2summarizes the univariate analysis for the complete study cohort. Multivariate analysis identified six variables as independent predictors of death or AMI (Table 3).These were older age (p < 0.00001), previous MI (p = 0.001), diabetes (p < 0.00001), pharmacologic stress (p < 0.00001), stress LV volume (p < 0.00001), and sLHR (p < 0.00001). The relative hazard increased by a factor of 1.5 (95% CI 1.2 to 1.8) for each standard deviation (SD) increase in sLHR after adjustment for all other variables.

A secondary analysis was performed in those 556 patients undergoing treadmill exercise only (Table 4).Multivariate analysis identified five variables as independent predictors of an adverse outcome (Table 5).These were male gender (p = 0.01), diabetes (p = 0.006), peak workload (p = 0.00006), stress LV volume (p = 0.0001), and sLHR (p = 0.003). The relative hazard increased by a factor of 1.4 (95% CI 1.1 to 1.8) for each SD increase in sLHR after adjustment for all other variables.

Incremental contribution

Figure 3shows the incremental improvement in the global chi-square for predicting death or AMI, using variables that were found to be independent predictors of outcome. As no perfusion variables were in these final models, we also included SSS. There was a statistically significant increase when the model using clinical variables alone (global chi-square 92.1) was augmented with variables related to the stress test (122.0, p < 0.00001), perfusion data (127.0, p = 0.08), left volume (150.7, p < 0.00001), and lung uptake (168.6, p < 0.00001). In the exercise-only subgroup, the inclusion of sLHR also provided a small but significant incremental improvement in risk stratification when added to clinical, stress test, perfusion, and volume information (global chi-square 85.5 vs. 77.1, p = 0.004).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Incremental contribution of clinical, stress procedure, perfusion, left ventricular volume, and lung uptake variables on prediction of acute myocardial infarction (MI) or death (see text for details). (A)All patients. (B)Treadmill exercise patients only.

Lung uptake measurements

The LHR measurements increased in association with worsening stress perfusion abnormalities. Lung uptake of ²⁰¹Tl after stress has been correlated with a rise in pulmonary capillary wedge pressure due to exercise-induced LV dysfunction (12). Increased resting pulmonary ²⁰¹Tl uptake has also been correlated with LV dysfunction, the extent of CAD, and resting perfusion abnormalities (13,14). Hurwitz et al. (15) reported that lung sestamibi was elevated in 15% to 17% of exercise studies using planar images obtained 4 min after injection, versus 9% to 10% on delayed acquisitions. Increased values were related to ischemic and fixed perfusion defects, LV dilation, LV dysfunction, and coronary stenoses. This suggests that the mechanism for enhanced ^99mTc-sestamibi lung uptake is similar to lung uptake of ²⁰¹Tl. The disadvantage of early imaging is that it negates much of the convenience and scheduling efficiency of sestamibi over thallium. A direct comparison of the relative prognostic value of early lung uptake characterization and delayed imaging is needed.

Survival analysis

The sLHR and rLHR were significantly higher in individuals who experienced AMI or death than in those individuals who remained event-free. The degree of univariate risk stratification, as determined by ROC analysis, was similar for sLHR, rLHR, and SSS. In contrast, a change in lung uptake provided minimal risk stratification. Other reports have emphasized that prognostic variables may differ in terms of predicting AMI and death (16), but in our cohort, sLHR provided similar risk stratification for these individual end points.

The variables independently associated with a lower event-free survival were older age, previous MI, diabetes, use of pharmacologic stress, larger stress LV chamber size, and sLHR. In the exercise only subgroup, peak workload was also predictive. Surprisingly, perfusion, as assessed by SSS, was no longer a significant predictor after adjustment for all other information. This finding is consistent with the observation that increased lung thallium uptake was the strongest predictor of a cardiac event (relative risk 3.5, 95% CI 2.2 to 5.4) (1).

Incremental contribution

The sLHR was additive to a model containing clinical, stress, perfusion, and LV volume information. Although the incremental improvement was modest, LHR measurement is available without any additional cost. Because no attempt was made to standardize the time between tracer injection and scanning, there was no interference with optimal camera utilization.

Study limitations and clinical implications

Although elevated lung uptake of sestamibi (highest tertile) is a strong univariate predictor of death or AMI (annualized event rate 4.8% per year), uptake in the lowest tertile is still associated with an intermediate risk (annualized event rate 1.6% per year), and other clinical and laboratory information must be considered in clinical decision-making. A limitation of our study is the lack of information on LV performance. Prognostic information can be obtained from the gated SPECT measurement of LV ejection fraction when combined with an assessment of perfusion (17). Our use of ungated LV volume as a surrogate for LV ejection fraction is not ideal but can be justified from the strong inverse relationship between these two parameters (18). Reliance on administrative data for assessing end points could also be questioned, although the Manitoba Population Health Research Data Repository has been used to validate prediction rules of mortality after AMI, and there is unlikely to be disagreement over the occurrence of death (19).

Conclusions

The sLHR appears to be another adjunctive prognostic measure in patients with known or suspected CAD. Before this measurement is routinely used for clinical decision-making, further studies are required to confirm our finding that lung uptake measurements can enhance risk stratification over conventional prognostic variables, and these studies should include a direct measurement of LV ejection fraction.