Author + information
- Received March 15, 2010
- Accepted April 12, 2010
- Published online June 1, 2010.
- Alex Exuzides, PhD⁎,
- Michael L. Main, MD†⁎ (, )
- Chris Colby, PhD⁎,
- Paul A. Grayburn, MD‡,
- Steven B. Feinstein, MD§ and
- Jonathan H. Goldman, MD∥
- ↵⁎Reprint requests and correspondence:
Dr. Michael L. Main, Saint Luke's Mid America Heart Institute, 4330 Wornall Road, Suite 2000, Kansas City, Missouri 64111
Objectives To compare acute mortality in critically ill hospitalized patients undergoing echocardiography with and without an ultrasound contrast agent (UCA).
Background Because of serious cardiopulmonary reactions reported immediately after administration of perflutren-containing UCAs, the FDA required a black box safety warning for this class of agents, including perflutren protein-type A microspheres injectable suspension.
Methods This study used the largest hospital service-level database in the U.S. All adult patients undergoing in-patient echocardiography between January 2003 and October 2005 were identified (n = 2,588,722, of which 22,499 received perflutren protein-type A microspheres injectable suspension). Of the 22,499 contrast echocardiography patients, 2,900 had diagnoses meeting criteria for critical illness (heart failure, acute myocardial infarction, arrhythmia, respiratory failure, pulmonary embolism, emphysema, and pulmonary hypertension). To control for the differences between the contrast and noncontrast patients, we used propensity score matching. Variables used in the construction of the propensity score included comorbidities, demographic factors, hospital-specific factors, level of care, and mechanical ventilation status. Patients receiving contrast echocardiography were matched to 4 control patients who received noncontrast echocardiography. Conditional logistic regression was used to estimate mortality effects.
Results There were 167 deaths in the study among critically ill patients, 38 of 2,900 from the contrast group and 129 of 11,600 from the control group. The contrast agent was not associated with an increase in same-day mortality (odds ratio: 1.18; 95% confidence interval: 0.82 to 1.71; p = 0.37). Before matching, contrast patients showed greater morbidity than noncontrast patients (Deyo-Charlson comorbidity score 2.45 vs. 2.25, p < 0.0001). After propensity score matching, these differences were significantly reduced, showing that both groups were well balanced.
Conclusions There is no increase in mortality in critically ill patients undergoing echocardiography with the UCA compared with case-matched control patients.
In October 2007, the U.S. Food and Drug Administration (FDA) issued a new black box warning, multiple new disease state contraindications, and a mandated 30-min monitoring period after contrast agent injection for the perflutren-based ultrasound contrast agents (UCA) (Optison, perflutren protein type A microspheres for injectable suspension [GE Healthcare, Buckinghamshire, United Kingdom] and Definity, perflutren lipid microsphere injectable suspension [Lantheus Medical Imaging, North Billerica, Massachusetts]) (1). The actions of the FDA authorities appeared to coincide with the post-marketing reports of 11 deaths after UCA administration and approximately 200 cases of serious nonfatal reactions. These self-reported events included deaths and adverse events that were reported but not independently adjudicated. To place these events in context, these reports occurred over a period of approximately 10 years after nearly 3.5 million injections of these UCAs.
In July 2008, the FDA changed the contraindications to warnings for both UCAs. The modification of label was preceded by a grassroots advocacy group of clinicians (2,3), the publications of peer-reviewed data studies (4), and a realistic understanding of the probable risk–benefit relationship for contrast-enhanced echocardiography in patients (5–7,8).
Coincident with the modified labeling changes and before the convening of a Cardio-Renal Advisory Panel on UCA safety, the FDA officials announced that GE Healthcare and Lantheus Medical Imaging agreed to conduct 2 follow-up safety programs (9) that included the following: 1) a retrospective, observational study that would use an administrative database to compare acute mortality in critically ill hospitalized patients who had received UCA (similar in scope to a single-site published report ); and 2) a prospective, pulmonary artery hemodynamic study in which patients would receive intravenous injections of UCA during a diagnostic, cardiac catheterization. The objective of the current study was to compare same-day in-hospital mortality in critically ill patients undergoing echocardiography with the addition of UCA (Optison) with mortality in critically ill patients undergoing echocardiography without the use of a UCA. The secondary objective of this study was to compare in-hospital mortality on the same day as or the day after the echocardiography procedure.
This was a retrospective case–control analysis of mortality in critically ill patients undergoing echocardiography in the presence of UCA (Optison) contrast compared with matched control patients. The study design was prospectively approved by the FDA, and the study was performed in partial fulfillment of the risk mitigation plan. The analysis dataset was derived from Premier Perspective database (Premier, Inc., Charlotte, North Carolina). Premier uses a consistent and secure sample of over 500 hospitals in the U.S., with a 93% hospital retention rate. This database is the largest hospital-based, service-level comparative database in the U.S. It includes all billed items, including procedures, medications, laboratory, and diagnostic and therapeutic services (including drugs and devices). It also includes all diagnoses for each patient. Detailed service-level information was available for each hospital day and included medication information and location within hospital. Data collected included patient demographics (age, sex, and race/ethnicity) along with the hospital's teaching status, urban/rural location, and geographic region. The database was subjected to quality/integrity checks to ensure a high level of data reliability. A propensity-matching score matching approach was used for selection of noncontrast control patients.
The design of this study fulfilled the FDA officials' request for a retrospective comparison of the mortality rate among critically ill patients requiring UCA-enhanced echocardiography as compared with patients who did not receive UCA.
Differences in baseline and/or clinical characteristics among patients receiving UCA were accounted for by using a comparator group, those patients who did not receive UCA. This study design allowed for the control patients to be matched to cases with a similar propensity score.
Patient inclusion criteria
The population of interest comprised hospitalized patients from the database who underwent at least 1 clinically indicated echocardiogram during the period from January 1, 2003 to November 1, 2005, and who were at least 18 years old.
The final analysis sample included the subset of hospitalized patients who underwent echocardiography with UCA or without UCA. Both groups included critically ill patients as defined by the presence of least 1 of 6 unstable cardiopulmonary conditions: unstable heart failure (40.7% of cases), acute myocardial infarction (38.2%), arrhythmia (5.1%), respiratory failure (13.4%), pulmonary embolism or emphysema (2.5%), and pulmonary hypertension (0.2%) as the admitting diagnosis (Table 1). Critical illness was defined using the International Classification of Diseases, Clinical Modification, 9th edition (ICD-9-CM) codes.
Covariates included the available demographic and clinical variables for each patient listed in the study population. Additionally, hospital-specific variables included teaching status, urban/rural location, and geographic region. Also included were variables that indicated anticoagulant and nonsteroidal anti-inflammatory drug uses.
Individual patient baseline morbidity level was assessed using elements of the Deyo-modified Charlson comorbidity index (DM-CCI). The DM-CCI weighted patients by comorbid conditions to assess risk of mortality (10). The CCI was later modified by Deyo et al. (11), who matched the conditions to ICD-9-CM diagnosis codes. The DM-CCI method has been widely used as an adjustor to control for comorbidities in retrospective database analyses. Table 2 lists the initial covariates.
Propensity score matching was used to control for the differences in observable covariates and to reduce the selection bias in case–control studies (12). All analyses were performed using SAS 9.1.3 (SAS Institute, Inc., Cary, North Carolina).
The first step in the propensity score matching was to model the choice between Optison and noncontrast echocardiogram. Each of the variables listed in Table 2 was included in a logistic regression, and no interaction terms were included in this propensity score model. Stepwise regression analysis was used to select the covariates with the best predictive power. SAS PROC LOGISTIC with the SELECTION = STEPWISE option was used; the default significance level was 0.05.
For each patient who received UCA, 4 matched control patients were assigned who showed similar propensity scores. This selection process utilized the smallest difference between propensity scores among the contrast patients and the control population. This method is described as the nearest neighbor matching and was used to implement the SAS macro [greedy matching algorithm (13)]. This matching algorithm performed up to 7 passes to find a match for each case. First, it searched for control patients with propensity scores within a tolerance of 0.0000001 and progressively relaxes the tolerance by 1 digit until reaching a value of 0.1.
Of note, selection of more than 4 control patients per case did not result in a substantial gain in statistical efficacy (14). This process created clusters of closely matched individual patients and well-balanced case and control groups. To compare in-hospital mortality, a conditional logistic regression model was used and a clustering variable identified the 4 matched control patients for each contrast case (15). The adjusted odds ratio of mortality for the 2 groups was estimated along with a 95% confidence interval (CI). The propensity score matching adjusted for baseline imbalances between the 2 groups. No covariates were used in the conditional logistic regression model for the primary, secondary, or exploratory analysis.
For the secondary analysis objective, the multivariable analysis was repeated as described above to compare in-hospital mortality on the same day as or day after echocardiography.
There were no patients with missing data once all inclusion and exclusion criteria had been applied. There was no imputation for missing values.
There were 208,878 noncontrast patients and 2,900 contrast patients who met all of the eligibility criteria for the study. The differences in covariates before the propensity score matching are shown in Table 3. Of the 29 covariates, 25 show statistically significant differences between the contrast patients and the noncontrast patients (with 18 covariates with a value of p < 0.0001, including the DM-CCI score).
These differences included potentially important confounding variables such as chronic obstructive pulmonary disease, coronary artery disease, renal disease, and diabetes (both type 1 and type 2). There were differences in the level of care between the noncontrast and contrast patients. Specifically, 17.0% of contrast patients were treated with mechanical ventilation, compared with 11.0% of the noncontrast patients (p < 0.0001).
After matching on propensity score, the observable difference between the 2 groups was reduced (Table 4). Only 5 covariates showed statistically significant differences between the contrast patients and the matched control patients: northeast hospital region (p = 0.0032), south hospital region (p = 0.0155), anticoagulation therapy (p = 0.0138), nonsteroidal anti-inflammatory use drug (p = 0.0057), and renal disease (p = 0.0439). For contrast patients, 17.0% were treated with mechanical ventilation, compared with 17.1% of the matched control patients (p = 0.8860). The overall similarity of the 2 groups and the similarity in the clinically important predictors, such as DM-CCI and mechanical ventilation status, indicates that the 2 groups were well balanced (for additional information on propensity score model results, please see the Online Appendix).
The final set of variables selected by the stepwise regression is shown in Table 5. Results for the primary and secondary analyses are shown in Table 6. There were 167 same-day deaths in the study, 38 from the contrast group and 129 from the control group. There were no significant differences between the contrast patients and the matched control patients for same-day all-cause mortality (adjusted odds ratio: 1.184; 95% CI: 0.820 to 1.709; p = 0.3675). The C statistic was 0.706, indicating that the model had good discriminative power.
There were 267 same-day or next-day deaths in the study, 62 from the contrast group and 205 from the control group. There were no statistically significant differences between the contrast patients and the matched control patients for same-day or next-day all-cause mortality (adjusted odds ratio: 1.222; 95% CI: 0.913 to 1.637; p = 0.1780).
The same methodology was used to explore the effects of UCA on same-day mortality for 3 mutually exclusive subgroups: patients receiving mechanical ventilation, intensive care unit (ICU) patients without mechanical ventilation, and coronary care unit (CCU) patients without mechanical ventilation. These results are shown in Table 7. None of the subgroups showed a statistically significant difference in same-day mortality.
There were 108 same-day deaths in the mechanical ventilation subgroup, with 22 in the contrast group and 86 in the control group. The adjusted odds ratio was 0.832 (95% CI: 0.488 to 1.420; p = 0.5120).
There were 30 same-day deaths in the ICU subgroup, with 9 in the contrast group and 21 in the control group. The adjusted odds ratio was 2.168 (95% CI: 0.892 to 5.272; p = 0.0878). There were 11 same-day deaths in the CCU subgroup, with 3 in the contrast group and 8 in the control group. The adjusted odds ratio was 1.42 (95% CI: 0.376 to 5.361; p = 0.6052).
In October 2007, the FDA mandated significant product labeling changes for UCA after several investigator-reported deaths and approximately 200 “serious, non-fatal reactions” that were temporally related, although not adjudicated and, therefore, not clearly causally attributable to UCA administration (1). In July 2008, subsequent to publication of new safety data in hospitalized patients undergoing echocardiography with and without a UCA (4), physician advocacy (2,3), and an enhanced appreciation of UCA efficacy (5–7), the FDA officials again modified the product labels to reflect the safety profile of available UCA (8). Conditional and consistent with the labeling revisions, industrial houses preparing UCA developed a risk mitigation plan to further define the safety profile of UCA (9). Part of this plan included a retrospective observational study using an administrative database to compare acute mortality in critically ill hospitalized patients undergoing echocardiography with and without a UCA.
Two similarly designed study results were recently reported. The first retrospective study, Kusnetzky et al. (4), compared acute mortality in hospitalized patients undergoing echocardiography with UCA (Definity) (n = 6,196) to those who did not receive UCA (n = 12,475). Those patients who received contrast showed higher clinical acuity, and correspondingly, more comorbidity than patients undergoing unenhanced echocardiography. Although proportionally more critically ill patients were housed in the contrast group, there was no increase in 24-h mortality in patients receiving the UCA (mortality rate 0.42% in the contrast arm and 0.37% in the unenhanced echocardiography arm, p = 0.60).
In a second retrospective study that used the Premier Perspective database, Main et al. (16) evaluated the 1-day mortality in 4,300,966 hospitalized patients who underwent echocardiography with (n = 58,254) or without (n = 4,242,712) UCA (Definity). Crude mortality was similar in both groups (contrast mortality at 1 day = 1.06% vs. 1.08% for unenhanced studies, p = 0.613). Multivariate regression analysis revealed that patients who received contrast were 24% less likely to die within 1 day as compared with patients who underwent unenhanced echocardiography (odds ratio: 0.75, 95% CI: 0.70 to 0.82).
The results of the present study are consistent with the reports from 2 previous publications, such that: 1) the design of the current analysis was prospectively approved by FDA officials; 2) the study included only critically ill hospitalized patients; and 3) propensity score matching was used to control for the baseline differences in covariates and reduced the potential selection bias that may occur in case–control studies. Overall, there were no differences identified in same-day or next-day mortality in patients who received contrast at the time of echocardiography versus patients who did not receive contrast. In addition, a subgroup analysis included patients who received mechanical ventilation, ICU patients without mechanical ventilation, and CCU patients without mechanical ventilation. The results revealed that there was no same-day or next-day mortality difference. These results corroborate the established safety experience for UCA as reflected in the published literature for UCA. Of note, the recently established post-marketing safety record that began in 2008 now lists over 260,000 injections of either of the 2 available UCA (4,16–22).
This study has several limitations. First, the design is observational and treatment selection was nonrandom. Table 3 shows significant baseline differences between the contrast group and the noncontrast group. These differences were reduced after using a propensity-matching score. Second, the cause of death was not available, and only crude mortality was reported. Third, the discrete time of death was not available in the Premier database (only the day of death was recorded). The primary outcome in this study was all-cause, same-day mortality; the secondary outcome was all-cause, same-day or next-day mortality. The dataset listed mortality and procedures date, and did not include a specific time.
There was no increase in short-term mortality in patients who underwent contrast-enhanced echocardiography in a large, propensity-matched population of critically ill patients. These findings were confirmed in high-risk patient subsets that included individuals residing in ICUs and patients who required mechanical ventilation.
For results of the initial propensity score regression, please see the online version of this article.
Dr. Main has received research support from Lantheus Medical Imaging and General Electric Healthcare. Dr. Goldman is Chief Medical Officer of ICON Medical Imaging Dr. Grayburn has received research grants from Philips Ultrasound, Evalve, Inc., General Electric Healthcare, and the National Institutes of Health. Dr. Feinstein is a speaker for Abbott and Takeda and has received research grants from the National Institutes of Health and GE Healthcare.
- Abbreviations and Acronyms
- coronary care unit
- Deyo-modified Charlson comorbidity index
- U.S. Food and Drug Administration
- International Classification of Diseases, Clinical Modification, 9th edition
- intensive care unit
- ultrasound contrast agent
- Received March 15, 2010.
- Accepted April 12, 2010.
- American College of Cardiology Foundation
- U.S. Food and Drug Administration Alert
- Main M.L.,
- Goldman J.H.,
- Grayburn P.A.
- Kusnetzky L.L.,
- Khalid A.,
- Khumri T.M.,
- et al.
- Reilly J.P.,
- Tunick P.A.,
- Timmermans R.J.,
- et al.
- U.S. Food and Drug Administration Alert
- FDA Advisory Committee briefing document for the Cardiovascular and Renal Drugs Advisory Committee June 24, 2008: Safety Considerations in the Development of Ultrasound Contrast Agents
- Parson L.S.
- Schlesselman J.J.
- Wei K.,
- Mulvagh S.L.,
- Carson L.,
- et al.
- Dolan M.S.,
- Gala S.,
- Dodla S.,
- et al.
- Abdelmoneim S.S.,
- Bernier M.,
- Scott C.G.,
- et al.