Sunday, August 12, 2007

Relatively Normal: pediatric echo z-scores

Big people have big hearts. Little people have little hearts- and mitral valves, aortas, and pulmonary arteries. As a student sonographer, one of the few things I excelled at was committing things to memory. I was all about flashcards. I had flashcards for everything, but particularly for normal dimensions. Normal left atrium? 19-40mm. Normal IVS? 6-11. LV? 37-56. In the adult echo arena this knowledge served me well as part of a routine:
  • Perform study
  • Obtain standardized measurements
  • Compare measurements to established criteria
Executing these simple steps allowed me to declare some exams within normal limits and others, unequivocally abnormal. I basked in the power and the glory. But pediatric echo brings a whole new set of flashcards. There are all the same adult values, and then again several times over: for kids with BSA up to 0.5 m2, 0.5 to 1.0 m2, 1-1.5, and the babies, and preemies, and what about the transverse arch... Yikes. Too many flashcards.
The problem of relating normal values over a wide variety of body sizes is at least as old as pediatric cardiology itself. Knowing if a newborn's tricuspid valve is too small or a two year-old's coronary artery is too big, is essential to the modern practice of pediatric echo.Enter the nomogram: Found in the backs of textbooks and throughout the literature, nomograms soon became the staple that replaced my flashcards. Now, all I needed was a copy machine and notebook. Nomograms aren't perfect, however. If the nomogram is printed too small (as I have deviously done with the above sample) it is quite difficult to resolve small differences between our measurement and the printed reference- and almost all of them require some interpolation on our part. Still, for the most part, the power and the glory had returned. Soon however, the absurdity of sitting in front of a computer/echo reading station, while interpolating hash-marks in a notebook caught up with me. This is the Age of Information? Besides- what if an unscrupulous cardiology fellow absconded with our Precious Notebook of Nomograms? What if?
Underlying each nomogram is the theory that a predictable relationship exists between the independent measure (age, weight, BSA, etc.) and a dependent variable (coronary artery diameter, annulus dimension, etc.). Further underlying each of these relationships, is the assumption that these variables have a normal distribution in our population.When we perform an echocardiogram and measure, for instance, the left coronary artery, and then ask "is it normal?"-- what we are really asking is: "how does our measurement compare to the mean of the population of other (normal) humans of this size?" The answer is best given with one number: the z-score. The z-score tells us in one simple, elegant number how our measure relates to the population.
  • Exactly normal: z = 0
  • pretty much normal: ± 1
  • too small: -3
  • gigantic: +7
Provided the authors of all these nomograms have actually published the predictions equations themselves (and not just the nomograms), we can now extract the information, leverage it to construct the predicted mean and standard deviation for our measure, and report a z-score. No more flashcards. No more notebooks. Power. Glory.