Epidural test dose and intravascular injection in obstetrics: sensitivity, specificity.

The recent study of the use of epinephrine as a chronotropic test dose deserves some comment (1). The investigators gave 0, 5, 10, or 15 pg of epinephrine intravenously to healthy laboring parturients. Using a pulse oximeter, they monitored the maternal heart rate response. They found that both 10 and 15 pg of epinephrine incited an increase in maternal heart rate within 60 s of injection. They then retrospectively determined that all women had at least a 10-beats/min increase in pulse after epinephrine, 15 pg. Without further prospective evaluation of this criterion, they concluded that epinephrine, 15 pg, was 100% sensitive (i.e., it detected all intravenous [IV] epinephrine injections) as an epidural test dose. Four of 15 women receiving normal saline injection also experienced a 10-beats/min tachycardia (73% specificity). They then concluded that "(a) a test dose containing . . .15 pg of epinephrine is 100% sensitive as a marker of intravascular injection . ..; (b) the positive predictive value of this technique may range between 55% and 73%; and (c) 27%45% of epidural catheters may be unnecessarily removed in the presence of a positive test" (1). We want to make several points about the authors' methods and conclusions. First, they have confirmed our results that tachycardia is not a specific finding in laboring women (2). A test with a 25%-45% false positive rate is clinically unacceptable. Second, they did not prove their claim that "a test dose containing . . . 15 pg of epinephrine is 100% sensitive as a marker of intravascular injection." They based this conclusion on a retrospectively derived end-point (tachycardia > 10 beats/min within 60 s of injection). They offer no prospective data to confirm the reliability of this criterion. (In our study, one woman had no tachycardia after IV injection of epinephrine, 15 pg (2). Finally, they erred when extrapolating the results of their small study sample to the population of laboring women. Because they encountered no false negative tests does not imply that the true incidence of false negative test is zero (100% sensitive). The mathematics of the above statement is as follows. A population of women having a negative test dose result is made up of true and false negative results. If R is the incidence of false negative tests, then (1 R ) is the fraction of a population having true negative tests, and is the probability that a negative test result observed in a single patient will be a true negative. The probability ( P ) of true negative results in n consecutive patients is (1 R)". Or,