Dr Philipp and her colleagues propose an elegant screening tool to select women with menorrhagia for hemostatic testing. They must be commended for their work on the improvement of care for these women.

However, we have 2 major concerns with the current study.

The first is that the population does not leave much room for efficient and effective screening: with a pretest probability of any coagulation disorder of 71%, the use of simply testing everyone is already very high. Also, the unusually high prevalence of coagulation disorders could limit the generalizability of findings from this population.

The second is their interpretation of the screening tool as useful. Given the prevalence (pretest probability) of 71% coagulation defects in these women, a positive predictive value of 72% with a positive screening test does not add additional information. The tool seems to have simply selected a random sample of their population, with the same prevalence as in the total population. By adding high PBAC score or low serum ferritin, the size of the sample increases but it is still random, as illustrated by no change in the positive predictive value. To illustrate, if we would increase the sample to all women, sensitivity would further increase to 100%, with the constant predictive value of 71%. Another way to illustrate this is to calculate the positive likelihood ratio (LR), the factor that converts pretest probability to posttest probability if a test is positive. The major advantage of a LR is that it is independent of the disease prevalence in a given population, in contrast with the positive predictive value. A test that adds no information has a LR of 1. Tests with LR values between 0.5 and 2.0 are generally considered not useful. The screening tool in the Phillip study has a positive likelihood ratio of 1.06.

To test whether the tool might perform better in a population with a lower prevalence of coagulation defects, we applied it to a population at our clinic (unpublished data). In this group, platelet aggregation or coagulation defects were diagnosed in 29% of women. In this study, sensitivity was 67%, specificity 24%, positive predictive value was 27%, and positive LR was 0.87.

In conclusion, we cannot agree with Dr Philipp that, in its current form, the proposed screening tool is useful in clinical practice.