In our final iconoclasm series post, Peter Sasieni explores the merits of making cancer screening non-binary.
The classic paradigm for diagnostic testing and medical screening is that both the health-state and the test-result are binary. Tested individuals can be categorised in a 2×2 table as either healthy or diseased and testing either positive (abnormal) or negative (normal). It is based on this paradigm that the familiar terms: sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) are defined. As we have seen, such dichotomisation over simplifies the real world.
Sometimes test results can be expressed as a continuous score, for instance one might measure the level of haemoglobin in faeces. In such circumstances it is common to draw a “receiver operating characteristics” (ROC) curve and to measure the area under the curve (AUC). The terminology comes from the fact that ROC curves were first used with radar during World War II. In an ROC curve one considers each possible value of the test as a threshold – at or above that threshold the test is positive; below it is negative. One then plots the sensitivity at the threshold against “one minus the specificity”. The curve is produced by adding one point at (0,0) (formed by saying that all results are negative) and another at (1,1) (formed by saying that all results are positive) and joining up all the points. One problem with the ROC curve is that it is still focused on binary classification. Ultimately it is assumed that one will either declare the result to be positive or negative. There is no accommodation for allowing different management depending on whether the result is strongly positive, weakly positive, weakly negative or strongly negative.
When it comes to antenatal screening, the focus has been on PPV (expressed in terms of the risk of Down’s syndrome) and sensitivity (the proportion of all Down’s syndrome pregnancies identified). However, screening is based on a variety of factors (including maternal age) and tests. Although ultimately, the mother needs to make a binary decision to terminate the pregnancy or not, each woman can set her own specific risk threshold. Whereas the aim of screening is to make the risk as close as possible to either zero or one, decisions also need to be made as to how many tests to carry out. All tests have a cost associated with them and some also have a risk of causing pregnancy loss. In theory each woman could choose at what level of risk (close to zero) she would be happy not to have any further tests, and at what level (close to one) she would have an abortion without further testing.
This description of antenatal testing may seem non-binary but, really, one is still making a binary decision and considering a binary health status. But we know that not all Down’s syndrome pregnancies are the same. Some will result in miscarriage (so having an abortion just terminates the pregnancy slightly earlier than would have happened naturally). All children with Down’s syndrome have some degree of learning disability and delayed development, however some have very few health problems as a result of their condition. Others will need extra medical care and attention.
The views expressed are those of the author. Posting of the blog does not signify that the Cancer Prevention Group endorse those views or opinions.