Eric Lim

Composite Outcomes in Cardiovascular Research: A Survey of Randomized Trials

Context Many trials combine several events into a single composite outcome when reporting the effect of an intervention. Contribution Of cardiovascular trials published over 7 years, 37% used composite outcomes. The individual events making up the composite differed in their clinical significance and in their influence on the composite estimate of effect. Caution The findings may not apply to trials in other specialties. Implication Composite outcomes are used commonly in cardiovascular trials and rely on end points that vary in their clinical significance. The overall estimate of effect for a composite measure cannot be assumed to apply equally to each individual outcome. The Editors Clinical trials often use composite outcomes to provide an overall estimate of the effect of an intervention. Composite outcomes are appropriate under many circumstances, such as when no single most important outcome exists, when several end points are thought to be important, and when 2 end points are clinically related (such as nonfatal myocardial infarction and death from fatal myocardial infarction). Also, because death is an uncommon occurrence, very large sample sizes are required to detect differences in deaths among trial groups. Composite outcomes are thought to facilitate the evaluation of the effects of treatment on infrequent outcomes, such as death in smaller trials, and are a convenient way to represent a broader range of the beneficial effects of an intervention (1, 2). Composite outcomes, however, require that the constituent outcomes are clinically meaningful and interpretable (3). Ideally, those individual outcomes should share the same biological effect: Each should contribute equally to the composite, and patients should attach a similar importance to each individual outcome (4). In addition, clinicians and statisticians have noted that composite outcomes are difficult to interpret when results are inconsistent for the individual outcomes, composite outcomes often overstate the results of the individual outcomes, and composite outcomes are used even if each individual event is not considered equally important (47). To better understand these issues, we reviewed published reports of randomized trials in cardiovascular medicine and surgery to assess how composite outcomes are currently used, constructed, and reported and to quantitatively assess the contribution of common, clinically important individual outcomes to the composite outcome. Methods Study Identification To identify published cardiovascular clinical trials, we hand-searched each issue of 14 leading general medical, cardiology, and cardiothoracic surgery journals (New England Journal of Medicine, Lancet, JAMA, BMJ, Archives of Internal Medicine, Annals of Internal Medicine, Circulation, Journal of the American College of Cardiology, American Heart Journal, Heart, European Heart Journal, Journal of Thoracic and Cardiovascular Surgery, Annals of Thoracic Surgery, and European Journal of Cardiothoracic Surgery) from 1 January 2000 to 1 January 2007. We identified a cardiovascular trial on the basis of any reported drug treatment or mechanical intervention aimed at improving myocardial, heart valve, or coronary artery function. We accepted any trial that was described as randomized by the authors as randomized. A composite outcome was any outcome that combined 2 or more different outcome measures. Death as an outcome could refer to all-cause mortality or cardiovascular mortality. Data Abstraction We identified all potentially relevant abstracts from hand-searched journals and retrieved full reports of manuscripts for review. We included all 2-group, parallel-design trials and excluded studies with more than 2 groups and those that did not provide a breakdown of the composite outcome (by not specifying the individual outcomes or by reporting the composite outcomes in a time-to-event analysis). Four researchers abstracted the data, and we resolved differences by consensus. We compared the number of composite outcome events with the number of events for each individual outcome for each trial. When composite outcomes were used for both primary and secondary outcomes, we included the primary composite outcome only so that each trial contributed only 1 composite outcome to the analysis. When the same group of composite outcomes was reported more than once, we used results from the longest reporting interval. When more than 1 composite outcome was reported, we used the composite outcome with the largest number of component outcomes. Because more than 1 end point can occur in a single patient and occasionally the same end point can occur more than once, we screened each article to determine whether each end point was counted as a separate event or whether a hierarchy of end points was used. A study with a hierarchy of end points would contribute a single event for each patient, either the first to occur or the most important. Statistical Analysis We measured how frequently composite outcomes are used in cardiovascular trials and how frequently specific individual outcomes are used and included as part of a composite outcome. We also quantified the number of individual outcomes used in the composite, and the proportion of the composite outcome contributed by each individual outcome. We obtained bias-corrected and accelerated CIs by using bootstrapping with 1000 replications for these descriptive data. We used linear regression to ascertain the relationship between study sample size (dependent variable) and the number of end points (explanatory variable). For outcomes of death and revascularization, we determined how the individual outcome influences the estimated strength of association between the intervention and outcome. We did this by creating a loglog, inverse variance, weighted circle plot of the estimated relative risk for the composite outcome against the estimated relative risk for the composite outcome without death or revascularization as an individual outcome. The larger circles in those plots represent studies in which the number of events is greater and the variance is smaller, usually resulting from estimates derived from larger sample sizes. If the relative risk for the composite outcome with and without the end point (death or revascularization) remained exactly the same, the circle would fall on a diagonal line from the bottom left to the top right of the plot. We arbitrarily defined a clinically meaningful difference of 20% in the relative risk estimate between the composite outcome and the composite outcome without death or revascularization, and we illustrate this 20% threshold with 2 dashed lines on either side of the diagonal line. Circles outside this boundary represent a clinically meaningful difference in risk with and without the individual outcome (death or revascularization), and larger circles outside this boundary represent clinically meaningful differences with greater certainty. Larger circles toward the top left and bottom right quadrants signify a greater contribution of the individual outcome (death or revascularization) to a trials conclusions based on the composite outcome. We used Stata, version 9.0 (StataCorp, College Station, Texas), and R, version 2.4.1 (R Foundation for Statistical Computing, Vienna, Austria) for statistical analyses. Role of the Funding Source No funding was received for this study. Results We identified 1231 randomized cardiovascular trials published between 1 January 2000 and 1 January 2007, of which 454 (37%) used composite end points. We excluded 150 trials that did not have a 2-group, parallel design or did not specify individual outcomes, which left 304 trials (28%): 221 (73%) that used a composite outcome as a primary outcome and 83 (27%) that used a composite outcome for its secondary outcome only. A median of 3 (interquartile range, 3 to 4) individual outcomes were used by researchers to create their composite outcome. The Table shows the distribution and contribution of individual outcomes to the trials composite outcome. Death was the most common end point in the composite outcome in 298 (98%) trials. Table. Frequency and Contribution of End Points Consistent with the argument that the smaller the sample size of a trial, the more end points are included in the composite outcome to inflate the number of events, our linear regression analysis suggested that the inclusion of each additional individual end point to the composite was associated with 721 fewer trial participants (P= 0.071). The total number of events for the individual outcomes was more than the total number of events represented by the composite outcome in 222 (73%) trials and was fewer than the total number of events represented by the composite outcome in 17 (6%) trials. We could not determine whether the discrepancy in the former instance was attributable to the use of a hierarchy (that is, in which only 1 event in a patient who had 2 or more events was selected for inclusion in the composite outcome) because most trials did not report whether they used a hierarchy. A plot of the P values of the composite outcome revealed marked asymmetry around the midpoint of the plot (at P= 0.050) (Figure 1), suggesting possible publication bias or that individual outcomes were selected for inclusion in the composite outcome to ensure statistical significance for the composite. Figure 1. P values of the composite outcome. A symmetrical plot of the P values of the composite outcome, truncated between <0.001 and 0.100, illustrates the asymmetrical distribution of the points that lie predominantly below the dotted line, where P= 0.050. Death and revascularization differed in their contribution to the composite outcomes. In Figure 2, the circles scattered away from the dashed parallel lines toward the top left and bottom right quadrants are few and small, suggesting that differences in estimates of relative risk based on composite outcomes wi

Modified technique for mitral repair without ring annuloplasty

Mitral valve repair is the procedure of choice to correct mitral regurgitation. Most operative techniques use an annuloplasty ring to provide stability and durability to the correction. We present a modification of existing repair techniques, without the use of an annuloplasty ring, in which plication sutures allow both annular remodeling and stability. Clinical and echocardiographic follow-up in our series of 60 patients with a mean follow-up of 29 months is presented.

Surgery for left ventricular free wall rupture: patch glue repair without extracorporeal circulation.

Left ventricular free wall rupture is a dramatic complication after myocardial infarction. We present our experience with the simple, expedient technique of patch glue repair without extracorporeal circulation. Access is obtained via median sternotomy. Evacuation of blood and haematoma is undertaken and a Goretex patch exceeding the size of infarct is fashioned. The patch is applied to the epicardium using enbrucrilate surgical glue instilled with gentle pressure against the beating heart. We performed this technique on 17 patients from 1993 to 2001, with a operative (30-day) mortality of 23.5% with a post-discharge survival of 85% at 2.2 years.

Pyrexia after cardiac surgery: natural history and association with infection.

BACKGROUNDnPyrexia is common after major surgery, and infection is often an important consideration. To investigate the natural history and association with infection, we performed a prospective observational study.nnnMETHODSnFrom November 2000 to January 2001, we studied 219 patients undergoing cardiac surgery screening daily for wound, respiratory, urinary tract, and other infections. Pyrexia was defined as temperature above 37.5 degrees C.nnnRESULTSnOf 219 patients, 7 intraoperative deaths occurred and 1 patient was excluded because of preoperative endocarditis, leaving 211. The mean age (SD) was 64 (10) years, consisting of 172 male patients (81.5%). The proportion pyrexial on days 1, 2, and 5 was 30.0%, 25.8%, and 10.3%, respectively. More patients undergoing urgent or emergency procedures (17.7% versus 7.8%; P =.03) subsequently developed pyrexia. However, there were no differences in wound infection (3.4% versus 8.3%; P =.13), positive cultures for respiratory (14.7% versus 11.4%; P =.16), urinary tract (5.2% versus 2.0%; P =.09), or other infection (8.6% versus 7.3%; P =.71) in patients experiencing postoperative pyrexia compared with those who did not.nnnCONCLUSIONSnPyrexia is common after cardiac surgery and resolves in the majority of patients by day 5. Because there is no association between early pyrexia and infection, diagnosis of early postoperative infection by pyrexia alone is insufficient and is better established by clinical assessment with microbiological evidence.

A simple model to predict coronary disease in patients undergoing operation for mitral regurgitation

BACKGROUNDnCoexistent coronary disease can be identified in a third of patients with mitral valve disease. This study aims to evaluate candidate selection strategy using risk factor identification and logistic regression and to develop an additive model for the prediction of coexistent coronary disease.nnnMETHODSnThe sample is a consecutive series of patients who had mitral repair from 1987 to 1999. Sensitivities and specificities were calculated for each risk factor. Variables for prediction of coronary disease were entered into a univariate analysis, and predictors were entered into a forward and backward stepwise multivariate logistic regression model to form a predictive score. An additive model was derived from transformation of the logistic model. Receiver operating characteristic curves were used to compare discrimination and precision quantified by the Hosmer-Lemeshow statistic.nnnRESULTSnThe American Heart Association and American College of Cardiology risk factor identification selection criteria for the 359 patients who had screening coronary angiography yielded 100% sensitivity and 1% specificity. Risk prediction with our logistic model produced a receiver operating characteristic curve area of 0.91 and Hosmer-Lemeshow score of 3.4 (p = 0.9). Similar discriminating ability for our patients was achieved by the Cleveland Clinic logistic model (receiver operator characteristic curve area of 0.79; Hosmer-Lemeshow score of 12; p = 0.1). Our five-item additive model produced receiver operating characteristic curve area of 0.91 and Hosmer-Lemeshow score of 3.81 (p = 0.80).nnnCONCLUSIONSnSimple risk factor identification has excellent sensitivity but is limited by specificity. Logistic regression modeling is an accurate risk prediction method but is difficult to apply at the bedside. Simplicity and accuracy may be achieved by the logistic regression-derived simple additive model.

Biological efficacy of low versus medium dose aspirin after coronary surgery: results from a randomized trial [NCT00262275]

BackgroundThe beneficial effect of aspirin after coronary surgery is established; however, a recent study reported the inability of low doses (100 mg) to inhibit postoperative platelet function. We conducted a double-blind randomised trial to establish the efficacy of low dose aspirin and to compare it against medium dose aspirin.MethodsPatients undergoing coronary surgery were invited to participate and consenting patients were randomised to 100 mg or 325 mg of aspirin daily for 5 days. Our primary outcome was the difference in platelet aggregation (day 5 – baseline) using 1 μg/ml of collagen. Secondary outcomes were differences in EC50 of collagen, ADP and epinephrine (assessed using the technique of Born).ResultsFrom September 2002 to April 2004, 72 patients were randomised; 3 patients discontinued, leaving 35 and 34 in the low and medium dose aspirin arms respectively. The mean aggregation (using 1.1 μg/ml of collagen) was reduced in both the medium and low dose aspirin arms by 37% and 36% respectively. The baseline adjusted difference (low – medium) was 6% (95% CI -3 to 14; p = 0.19). The directions of the results for the differences in EC50 (low – medium) were consistent for collagen, ADP and epinephrine at -0.07 (-0.53 to 0.40), -0.08 (-0.28 to 0.11) and -4.41 (-10.56 to 1.72) respectively, but none were statistically significant.ConclusionContrary to recent findings, low dose aspirin is effective and medium dose aspirin did not prove superior for inhibiting platelet aggregation after coronary surgery.

A Method for Descending Thoracic Aortic Replacement Retaining a Posterior Strip Bearing Intercostal Vessels

Operations for aneurysms of the descending thoracic aorta are still fraught with danger. Spinal cord injury remains a major cause of morbidity. Many therapeutic strategies have been suggested to reduce the incidence of this devastating complication, including reimplantation of intercostal vessels. However, reimplantation of intercostal vessels, both individually or in groups, is time consuming and compounded by the absence of a reliable means of identifying which vessels actually supply the cord. We present a technique that allowed inclusion of all potentially important descending aortic branching vessels into the repair leading to a favorable outcome in a series of patients.

Candidate selection for heart transplantation in the 21st Century

Cardiac failure is an important problem in developed countries with impacts on patient prognosis, quality of life, and economic cost. Although heart transplantation is the only established therapy for advanced heart failure, it is limited by donor organ shortage. Medical therapy should be optimized before transplant assessment, and patients should be offered the option of alternative surgical therapy when possible. Appropriate candidate selection by precise risk stratification and exclusion of patients with unacceptable risk will ensure best utilization of a scarce resource to patients who will derive the most benefit.

Determinants of Successful Endoscopic Internal Thoracic Artery Harvesting: A Prospective Analysis

OBJECTIVEnEndoscopic internal thoracic artery (ITA) harvesting is employed during minimal-access coronary artery bypass grafting. To improve case selection, we prospectively analyzed our entire experience to identify variables that predict intraoperative conversion to sternotomy.nnnMETHODSnWe performed a prospective study from September 1999 to November 2003 of 100 consecutive patients with an endoscopically harvested left ITA (LITA). Success was defined as an endoscopic dissection of the LITA sufficient to reach the anastomosis. Multivariate logistic regression analysis was performed to identify independent preoperative and procedural predictors of success.nnnRESULTSnThe measured parameters (mean +/- SD) were age (62 +/- 9 years), height (174 +/- 9 cm), weight (81 +/- 14 kg), and logistic Euroscore (2.0 +/- 1.7). Patients comprised 8 (8%) women, 17 (17%) with urgent operations, 42 (42%) with multiple vessel disease, 17 (17%) with a left ventricular ejection fraction <50%, 2 (2%) redo procedures, and 3 (3%) with pleural disease. The Zeus robot was used in 17 patients (17%). Eight-eight (88%) of the LITA were successfully harvested endoscopically. Among the 12 patients who underwent conversions, pleural adhesions were the most common finding (n = 4, 33%). One LITA was unusable. In the final multivariate model, lung disease was a negative predictor of successful endoscopic harvest (odds ratio, 0.13; 95% confidence interval, 0.02-0.63; P =.012). The variables of age, sex, left ventricular function, logistic Euroscore, operative priority, and use of the Zeus robot did not achieve statistical significance.nnnCONCLUSIONSnAcceptable conversion rates and low conduit wastage are achievable during a units initial experience. Lung disease is associated with increased conversion frequency, and surgeons embarking on endoscopic harvesting should consider excluding these patients to improve their chances of success. Pleural adhesions increase the technical difficulty of surgery.