Aims To examine differences in patient characteristics and outcomes in 19 636 patients enrolled in the USA and 3027 patients enrolled in other countries undergoing intra-aortic balloon pump (IABP) counterpulsation.
Methods and results Indications for IABP use; a larger percentage of US patients were identified as ‘early support and stabilization for angiography or angioplasty’ (21.1% US vs 11.8% non-US), and ‘pre-operative support for high-risk CABG’ (15.9% vs 6.6%). A smaller percentage of US patients vs non-US patients were identified as ‘weaning from cardiopulmonary bypass’ (14.3% vs 28.2%), and ‘refractory ventricular failure’ (6.2% vs 9.8%). One out of five patients in both groups was listed as ‘cardiogenic shock’ (18.9% US vs 20.2% non-US). All cause, risk-adjusted, in-hospital mortality (20.1% vs 28.7%; P<0.001), and mortality with IABP in place (10.8% vs 18.0%; P<0.001) were lower at US vs non-US sites. In both US and non-US institutions, IABP associated complication rates, such as IABP-related mortality (0.05% vs 0.07%), major limb ischaemia (0.9% vs 0.8%), and severe bleeding (0.9% vs 0.8%), were low.
Conclusions IABP counterpulsation is deployed at an earlier clinical stage in US patients. Mortality rates are higher for non-US patients, particularly for patients with non-surgery cardiac interventions, even after adjusting for risk factors. Complication rates were low. Physicians should therefore not be reluctant to use IABP in high-risk patients undergoing cardiac procedures.
Heart assist device
Peripheral vascular disease
International registries are important tools for analysing disease management and outcomes among different nations. Exploration of such data may lead to changes in national health care policies and disease management strategies.1International registries in the United States (US) and Europe have documented regional differences in patient care, and as discrepancies have been identified, have helped to change patient care.2Patient treatment can vary substantially from region to region, as can baseline patient characteristics. For example, even within Europe, there is substantial geographic variation in the treatment and evaluation of acute coronarysyndromes.3
↵aP values for patient age, weight, body surface area, and body mass index were obtained from a one-way comparison of the ranks within the combined sample. P values for all other characteristics were obtained using a chi-square test.
ProphylacticaIABC use: incidence by insertion yearb
↵a Prophylactic use is defined as balloon insertion before lastintervention.
↵b Comparisons made using chi-square test on unadjusted rates.
↵c In 1997 there were less than 100 international entries.
The Benchmark counterpulsation outcomes registry is a prospective registry of all patients who receive an intra-aortic balloon counterpulsation (IABP) at participating institutions. An overview of the outcomes4revealed that IABP complication rates are low, although all cause in-hospital mortality remains high, particularly in high-risk patients. The primary aim of the present study is to examine differences in baseline characteristics, patient management, and outcomes between patients in the US and outside the US undergoing IABP from June 1996 through August 2001.
The Benchmark Counterpulsation Outcomes Registry currently includes 185 US sites and 65 non-US sites that use intra-aortic balloons manufactured by Datascope Corporation, Fairfield NJ, USA. Of the 22 663 patients enrolled in the Benchmark registry, 19 636 (87%) were enrolled in the US, and 3027 (13%) were enrolled in 18 other countries (Appendix A). Any patient at a participating hospital who received an IABP was entered into the database. Patients were registered consecutively to avoid biased selection of cases and were followed to hospital discharge. Concomitant medications were left to the discretion of the treating physician. Information on data collection, handling, and validation has been previously published.4Briefly, all data were collected by nurses with oversight provided by site investigators. Three separate audits were used to validate the accuracy of the data.5Datascope Corporation provided funding for the database.
For quantitative variables, such as age, the Student's t-test was applied to the ranks of the observations within the combined sample, a method also known as a rank transformation. For categorical variables, P values were determined by using a chi-square test comparing the US distribution versus non-US distribution over the categories. Data on prophylactic use are based on IABP insertion before an intervention (not during or after an intervention). Cardiac interventions were categorized as surgical or non-surgical interventions. If patients had more than one intervention, then only the last intervention was used.
Multiple logistic regression was used to compare the US and non-US in-hospital mortality experience after adjusting for mortality risk factors. Factors were chosen using model-building procedures developed from previous work with the Benchmark Registry. Potential factors were screened and models were generated in a stepwise procedure. First, individual factors were screened using two-way crosstabulations of the factor categories with in-hospital mortality, or by determining whether mortality/non-mortality groups separate the means of continuous factors. Generally, only factors with a statistically significant association (P<0.05) were included in the full model. Second, factors chosen in the first step were entered in a full logistic regression model. The effects of collinearity were determined as a reduction of predictive value in the full model of a factor that individually had a strong apparent relationship with in-hospital mortality. Third, non-significant factors (P>0.05) were eliminated from the full model in the following order, first interactions, then the individual factors. If an interaction was statistically significant, then the individual factors making up the interaction were automatically retained in the model regardless of their individual significance. Fourth, likelihood comparisons to the full model and model-fitting diagnostics were used to judge whether predictive value was lost by removing factors and or interactions. Finally, candidate factors were fitted to a randomly selected subset of records and the resulting logistic regression model coefficients were used to discriminate the mortality/non-mortality subgroups from another disjoint subset of records by computing the predicted logits and measuring the resulting separation of the mortality/non-mortality subgroups.
The risk factors and models differed depending on whether or not patients had a surgical cardiac intervention. Factors that were common to both surgical cardiac and non-surgical cardiac predictor models were age >75 years, body surface area (BSA), initial pre-IABP systolic blood pressure, ejection fraction <30%, peripheral vascular disease, wait for IABP ⩾5 days after hospital admission, insertion year, indication for use listed as ‘cardiogenic shock’, ‘unstable refractory angina’, or ‘support and stabilization’. Additional factors in the model for patients with surgical cardiac interventions included gender, primary diagnosis=acute myocardial infarction (AMI), previous coronary artery bypass graft (CABG), indication for use ‘wean from cardiopulmonary bypass’, or ‘pre-operative high-risk CABG’. Additional factors in the model for patients with non-surgical cardiac interventions included history of diabetes, previous MI, and ⩾3 vessel disease. The observed US and non-US mortalities were then computed within predicted risk quartiles generated from these logistic regression models, and this provided a graphical summary of the mortality experience within patient subgroups having similar mortality risk. By subdividing the patients into groups according to the quartiles determined by their estimated risk of death and then comparing the observed mortalities within these risk quartiles, we had the opportunity of observing not just one US/non-US difference, but several differences depending on the patient risk. The formal adjustedcomparisons of mortalities were then accomplished by including a US/non-US factor in both the surgery and non-surgery logistic regression models to generate odds ratios from the factor coefficients. For this study, the criterion for statistical significance was P=0.05 and was not adjusted for multiple comparisons.
We analysed risk adjusted mortality rates after excluding all countries that contributed less than 98 cases (Appendix A) and found no change in the results. All analyses were performed using SAS, version 8, using the Windows 98 operating system for an IBM compatible PC.
The primary endpoints included major limb ischaemia, severe bleeding, IABP failure (balloon failure and balloon leak), and all cause in-hospital mortality. All cause in-hospital mortality was defined as mortality occurring from any cause during IABP or after IABP. A secondary endpoint was major IABP-related complications, defined as any major limb ischaemia, severe bleeding, IABP leak, or mortality directly attributed to IABP. Examples of reasons for death attributable to IABP include stroke and multiple emboli to gut with subsequent bowel infarction. Definitions of all endpoints have been previously published.4
A series of independent audits used to validate the registry revealed that virtually all IABP procedures at participating sites were included. In addition, accuracy rates were ⩾95% for most categorical variables and ⩾80% for quantitative variables that required a writtenanswer.5The high rates of agreement between registry data and patient records support the reliability of the data.
Baseline clinical characteristics for the total population, for US patients, and for non-US patients are shown in Table 1. More US patients were women, had previous CABG surgery, or diabetes, or had an ejection fraction <35%. US patients also tended to be slightly older, taller, heavier, and have a greater body surface area (BSA) and body mass index (BMI). In contrast, more non-US patients had a history of myocardial infarction (MI).
3.1 Indications and insertion characteristics
Due to the large number of patients, most of the differences between the US and non-US sites are statistically significant. Thus, we emphasize only the major differences. Overall, the indications and locations of insertion were significantly different (P<0.001). Specifically, a greater percentage of IABP use, at US compared to non-US centres, was to provide early support and stabilization of patients undergoing angiography and angioplasty (21.1% US vs 11.8% non-US), or pre-operative support for high-risk CABG (15.9% vs 6.6%). A smaller percentage of IABP use, at US compared to non-US centres, was for weaning from cardiopulmonary bypass (14.3% vs 28.2%), or refractory ventricular failure (6.2% vs 9.8%) (Table 2). Roughly similar percentages of patients in both groups had IABP for pre-operative support for high-risk general surgery (0.7% vs 0.6%), or for cardiogenic shock (18.9% vs 20.2%). The average length of stay (P<0.001), the number of days from insertion to discharge (P<0.001), and the number of days with IABP in place (P=0.008) were lower in US compared to non-US institutions. Time from hospital admission to IABP insertion was shorter for patients at US centres, with 13.6% of patients receiving IABP ⩾5 days from admission as compared with 17.1% at non-US centres. In the US, 65.2% of insertions were in the catheterization laboratory versus 39.4% in non-US facilities (Table 2). The majority of insertions in non-US sites occurred in the operating room. The insertion was sheathless in 16.6% of US cases and in 41.5% of non-US cases. Lastly, prophylactic IABP use, defined as balloon insertion before last intervention, has increased over time, and is higher in US institutions (31.3% US vs 20.3% non-US in 2001, Table 3).
3.2 Complications and mortality rates
Complication and mortality rates, regardless of location, were low, and in specific, major complication rates (IABP-related mortality, balloon failure, balloon leak, major limb ischaemia, and severe bleeding) were extremely low (Table 4). In contrast, overall mortality rates in this seriously ill population remain high at more than 20%. In the Benchmark registry, all cause unadjusted in-hospital mortality, and mortality with IABP in place, were significantly lower at US institutions (20.1% US vs 28.7% non-US, P<0.001), and (10.8% vs 18.0%, P<0.001) (Table 4). An investigation of all causein-hospital mortality and mortality with balloon in place by insertion year reveals that both are consistently higher in non-US institutions (Table 5).
In-hospital mortality and mortality with balloon in place—incidence by insertion yeara,b
Mortality: ballon in place
↵a Data from 1996 are not included because there were so few international entries in that year.
b Chi-square tests applied to unadjusted rates.
Since it is possible for mortality rates to be higher in non-US sites because patients were simply at higher risk and had more comorbidity, we adjusted for risk factors and analysed the data with multiple logistic regression analysis. After adjusting for risk factors, patients in non-US institutions were still at higher risk for mortality. Patient data were subsequently divided into groupsaccording to predicted mortality based on risk factors. For patients with non-surgical cardiac interventions, mortality in non-US sites was consistently higher in every risk group (Fig. 1). Similarly, mortality was higher in non-US sites for patients with surgical cardiac interventions, although the differences were not consistently significant (Fig. 2). Mortality was significantly higher for non-US patients with non-surgical cardiac interventions whose primary indication was cardiogenic shock (adjusted odds ratio, [95%CI]=1.54, [1.27, 1.87]; P<0.001) (Fig. 3).
In-hospital mortality by predicted risk for patients with non-surgical cardiac interventions. Error bars indicate 95% confidence intervals. The predicted mortality categories were based on a logistic regression model combining US and non-US sites. Category boundaries are quartiles of the predicted in-hospital mortalities from the combined cohort of non-surgical cardiac interventions.
In-hospital mortality by predicted risk for patients with surgical cardiac interventions. Error bars indicate 95% confidence intervals. The predicted mortality categories were based on a logistic regression model combining US and non-US sites. Category boundaries are quartiles of the predicted in-hospital mortalities from the combined cohort of surgical cardiac interventions.
In-hospital mortality by predicted risk by primary indications. CG=cardiogenic. Category boundaries are quartiles of the predicted in-hospital mortalities from the combined indications.
The logistic regression model that included a non-US versus US term revealed that non-US patients with surgical cardiac interventions had higher all cause in-hospital mortality (adjusted odds ratio, [95% confidence interval (CI)]=1.34, [1.21, 1.49]; P<0.001). Non-US patients with non-surgical cardiac interventions also had high all cause in-hospital mortality (adjusted odds ratio, [95% CI]=1.72, [1.51, 1.95]; P<0.001).
Mortality with balloon in place was also higher in non-US patients with both surgical cardiac interventions (adjusted odds ratio, [95% CI]=1.48, [1.31, 1.68]; P<.0.001) and with non-surgical cardiac interventions (adjusted odds ratio, [95% CI]=1.74, [1.50, 2.02]; P<0.001).
In US and non-US institutions, IABP is generally associated with a low rate of major complications. However, all cause in-hospital mortality and mortality with balloon in place are higher in non-US sites even after adjusting for risk factors. Potential reasons for this difference in mortality rates include the greater use of IABP in a prophylactic mode at US centres and at an earlier time point in the patients’ illness. Whereas the prophylactic use of IABP has also increased at non-US centres, corresponding reductions in mortality may have not been realized as the rate of IABP use for more indications in high-risk patients has also increased. IABP insertion location was more commonly in the catheterization laboratory in US centres and more commonly in the operating room in non-US centres. This difference in insertion location may reflect differences in indications for IABP use—non-US centres use IABPs for surgical cardiac interventions more frequently than US centres. Previous studies support the notion that having a catheterization laboratory can affect treatment approaches and outcomes. For example, the GRACE investigators found substantial differences in treatment approach according to teaching status, presence of catheterization laboratory, and geographic region.6,7However, the presence or absence of a catheterization laboratory in the GRACE study was not shown to affect mortality.7
Hospital length of stay is also longer in non-US sites. This may be a result of the selection of high-risk patients for IABP. Patients at non-US institutions may also receive their IABP later in the course of their treatment. Combined with lower rates of prophylactic IABP use at non-US sites, this may explain some of the observed differences in mortality rates. It is also possible that US centres may use IABP more commonly in patients who are destined to do well.
Many large international trials have demonstrated substantial variability in outcomes among different countries.8–10For example, after adjusting for baseline characteristics, enrolment in the United States was found to be a marginally significant predictor of better survival in the GUSTO I trial.8They also found that more complications were reported in the United States. Even in the United States, there is great variation in the risk-adjusted rates of IABP use between hospitals.11
Although IABP use has been shown to be cost-effective and beneficial in high-risk patients,12–15controversy persists about appropriate indications for use because of the historically high complication rates.16,17However, major complication rates in the Benchmark registry were small, ranging from 0.9% to 2.7%. Physicians should therefore have little reluctance to use IABP in high-risk patients undergoing cardiac procedures. For example, IABP treatment improves the outcomes of patients who have left-main coronary disease,18an ejection fraction of <0.25,19who are undergoing repeat CABG surgery,20or are undergoing cardiac catheterization during myocardial infarction.14However, Stone et al, demonstrated no difference in mortality with prophylactic treatment with IABP versus conservative treatment in patients with acute myocardial infarction undergoing percutaneous transluminal coronary angioplasty,21although the event rate was extremely low (4.3% vs 3.1%; n=437).
The limitations of this study are inherent to any large-scale post hoc analysis of prospectively gathered data. Because the registry contains a high number of patients, a difference of just 1% will likely have a significant P-value. Although the differences are unlikely to be attributed solely to chance, the clinical importance of the differences must be judged with practical wisdom and clinical experience. In addition, this is not a randomized trial; rather, it is a detailed description of ongoing and evolutionary clinical practice. There may also be site-to-site variations in personnel and resources allocated to the registry, individual practice patterns, and patient populations that are intrinsic to any multinational commercially-sponsored registry.
In the multiple logistic regression analysis, all possible risk factors for mortality were not included in the model. Variables that were not collected as part of the registry could not be included in the analysis.
Indications for IABP use differ between non-US institutions and US institutions. US institutions use IAPB for the early support and stabilization of patients undergoing angiography and angioplasty, pre-operative support for high-risk CABG, and cardiac support for high-risk surgery more frequently than their non-US counterparts. Non-US sites are more likely to use IABP in the treatment of high-risk patients including treatment of cardiogenic shock, weaning from cardiopulmonary bypass, and refractory ventricular failure. Overall, major complication rates are low and physicians should be less reluctant to use IABP in high-risk patients undergoing cardiac procedures. Prophylactic use of IABP has increased over time, and remains higher in US institutions. All cause in-hospital mortality and mortality with balloon in place are significantly higher for patients undergoing non-surgical interventions at non-US sites even after adjusting for risk factors.
Appendix A: Percentages do not total 100 due to rounding; three missing values for non-US sites
Number of patients
The Benchmark Registry and this analysis were supported by a grant from Datascope Corporation, Fairfield, New Jersey, USA.
FoxKAA, Cokkinos DV, Deckers J et al. The ENACT study: a pan-European survey of acute coronary syndromes. European Network for Acute Coronary Treatment. Eur Heart J. 2000;21:1440–1449 doi:10.1053/euhj.2000.2185.
FergusonJJrd, Cohen M, Freedman RJ et al. The Current Practice of Intra-Aortic Balloon Counterpulsation: Results Form the Benchmark Registry. J Am Coll Cardiol. 2001;38:1456–1462 doi:10.1016/S0735-1097(01)01553-4.
EagleKA, Goodman SG, Avezum A, GRACE Investigators et al. Practice variation and missed opportunities for reperfusion in ST-segment-elevation myocardial infarction: findings from the Global Registry of Acute Coronary Events (GRACE). Lancet. 2002;359:373–377 1016/S0140-6736(02)07595-5.
FoxKA, Goodman SG, Klein W et al. Management of acute coronary syndromes. Variations in practice and outcome. Findings from the Global Registry of Acute Coronary Events (GRACE). Eur Heart J. 2002;23:1177–1189 1053/euhj.2001.3081.
van de WerfF, Topol EJ, Lee KL et al. Variations in patient management and outcomes for acute myocardial infarction in the United States and other countries: Results from the GUSTO trial. Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries. JAMA. 1995;273:1586–1591.
BarbashGI, Modan M, Goldbourt U et al. Comparative case fatality analysis of the International Tissue Plasminogen Activator/Streptokinase Mortality Trial: variation by country beyond predictive profile. J Am Coll Cardiol. 1993;21:281–286.
OhmanEM, George BS, White CJ et al. Use of aortic counterpulsation to improve sustained coronary artery patency during acute myocardial infarction. Results of a randomized trial. The Randomized IABP Study Group. Circulation. 1994;90:792–799.
CohenM, Dawson MS, Kopistansky C et al. Sex and other predictors of intra-aortic balloon counterpulsation – related complications: Prospective study of 1119 consecutive patients. Am Heart J. 2000;139:282–287.
StoneGW, Marsalese D, Brodie BR et al. A prospective, randomized evaluation of prophylactic intraaortic balloon counterpulsation in high risk patients with acute myocardial infarction treated with primary angioplasty. Second Primary Angioplasty in Myocardial Infarction (PAMI-II) Trial Investigators. J Am Coll Cardiol. 1997;29:1459–1467.
MarcCohen, PhilipUrban, Jan T.Christenson, Debra L.Joseph, Robert J.FreedmanJr, Michael F.Miller, E.MagnusOhman, Ramachandra C.Reddy, Gregg W.Stone, James J.FergusonIIIEur Heart J(2003)24 (19):
1763-1770DOI: http://dx.doi.org/10.1016/j.ehj.2003.07.002First published online: 1 October 2003 (8 pages)