To the Editor.
—Although Dr Connors and colleagues1 claim that right heart catheterization (RHC) results in increased mortality, careful consideration of the data using the evidencebased medicine approach2 described in JAMA raises numerous concerns. First, the conclusions are drawn from subgroup analyses. The matched cohorts are subgroups of larger populations that are significantly different in determinants of mortality such as the APACHE III score, blood pressure, renal function, and pulmonary function. The post hoc matching process is therefore at risk for considerable bias and/or confounding. An example of this risk is seen in the use of the propensity score to define the cohorts. An acceptable match for an RHC patient was required to have a propensity score within 0.03 units of the RHC patient. This propensity score limit appears to have been arbitrarily determined. The sensitivity analysis the authors performed failed to consider possible selection bias resulting from
Baxter JK, Beilman GJ, Abrams JH. Effectiveness of Right Heart Catheterization: Time for a Randomized Trial. JAMA. 1997;277(2):108. doi:10.1001/jama.1997.03540260022012