Letter by Bouzas-Mosquera et al Regarding Article, “Cardiac Magnetic Resonance With T2-Weighted Imaging Improves Detection of Patients With Acute Coronary Syndrome in the Emergency Department”
To the Editor:
We read with interest the article by Cury et al1 on the value of cardiac magnetic resonance (CMR) with T2-weighted imaging for the detection of acute coronary syndrome (ACS) in patients presenting to the emergency department with acute chest pain. In this study, 62 patients were evaluated and 13 developed ACS. The authors concluded that the new CMR protocol significantly improved the diagnostic accuracy for the detection of ACS and added incremental value over clinical risk stratification and traditional cardiac risk factors (TCRFs). Thus, we commend the authors for their findings given the potential consequences of misdiagnosis in these patients. However, some results of this study may be questionable because of several methodological and statistical flaws.
First, a number of problems exist relative to the strength of the data on background knowledge. When the incremental value of imaging over pretest data is assessed, the use of an optimized model of pretest information is essential. A suboptimal model may result in overestimation of the incremental value of imaging.2 To characterize the pretest risk of the patients, the authors used a subjective scale provided by the emergency department caregivers. The use of a subjective assessment of pretest probability of ACS not only may undermine the predictive value of the clinical model but also precludes the replication of the results. On the other hand, for the TCRFs assessment, the algebraic sum of the number of TCRFs was used. However, this risk assessment may be inaccurate because the higher predictive value of diabetes mellitus and family history of myocardial infarction over other TCRFs3 was not considered.
Second, the authors stated that the addition of the new CMR protocol to the model containing the clinical risk assessment and TCRFs significantly improved the c-statistic from 0.771 to 0.958 (P<0.0001). Unfortunately, the confidence intervals of the areas under the receiver-operating-characteristic curves of the different models were not provided. Furthermore, the P value of <0.0001 might be misleading because it does not represent the P value of the difference between the areas under the 2 receiver-operating-characteristic curves, but the P value of the increase in the global χ2 of the model. The use of other methods for comparison4 would have been more appropriate.
Third, the authors claim that “subjects with a positive CMR had a 129-times higher likelihood of ACS than those with a normal CMR after adjustment for initial clinical risk assessment and TCRFs.” However, the estimated 95% confidence interval of the odds ratio ranged from 11.8 to >1000. When the events of interest are sparse, the precision of the estimated effects may be very low and the resulting regression coefficients may not be reliable.5
The performance of multivariate analysis in small data sets has important limitations, and the authors should acknowledge that a sample of 62 patients with only 13 events of interest is too small to allow for a trustworthy multivariable logistic regression analysis.
Cury RC, Shash K, Nagurney JT, Rosito G, Shapiro MD, Nomura CH, Abbara S, Bamberg F, Ferencik M, Schmidt EJ, Brown DF, Hoffmann U, Brady TJ. Cardiac magnetic resonance with T2-weighted imaging improves detection of patients with acute coronary syndrome in the emergency department. Circulation. 2008; 118: 837–844.
Hachamovitch R, Di Carli MF. Methods and limitations of assessing new noninvasive tests: part II: outcomes-based validation and reliability assessment of noninvasive testing. Circulation. 2008; 117: 2793–2801.