D-dimer Interval Likelihood Ratios for Pulmonary Embolism.

Abstract

The objective was to estimate D-dimer interval likelihood ratios (iLRs) for diagnosing pulmonary embolism (PE). The authors used pooled patient-level data from five PE diagnostic management studies to estimate iLRs for the eight D-dimer intervals with boundaries 250, 500, 750, 1,000, 1,500, 2,500, and 5,000 ng/mL. Logistic regression was used to fit the data so that an interval increase corresponds to increasing the likelihood ratio by a constant factor. The iLR for the D-dimer interval 1,000-1,499 ng/mL was essentially 1.0 (0.98 with 95% confidence interval [CI] = 0.82-1.18). In the logistic regression model, the constant between-interval factor was 2.0 (95% CI = 1.9-2.1). Using these iLR estimates, if the pre-D-dimer probability of PE is 15%, only a D-dimer less than 500 ng/mL will result in a posttest probability below 3%; if the pretest probability is 5%, the threshold for a "negative" D-dimer is 1,000 ng/mL. A decision strategy based on these approximate iLRs agrees with several published strategies.