Reply letter: Screening for thyroid disease in pregnancy

Sir,

We have read with much interest L. Valdivielso's letter, especially those comments relating to scientific aspects of our paper, which will undoubtedly contribute to increasing our knowledge regarding the costs of thyroid disease in pregnant Spanish women.

As regards methodology, this letter suggests that the accuracy and reliability of the results may be inadequate. Among other aspects, it is suggested that the decision model delineated in Fig. 1 apparently lacks the initial node. Further reading makes it clear that the first screening performed on the hypothetical cohort of pregnant women is based on the measurement of peroxidase antibodies (anti-TPO). A rapid and simple calculation between the two types of anti-TPO (0.11 + 0.89 = 1) makes it evident that the initial node would correspond to the total group of pregnant women considered.

On the other hand, the assignment of probabilities and utilities used (which, as noted in the letter by L. Valdivielso, have been widely used in previous articles) is also questioned. However, these two parameters are confused and interchangeably used in his reasoning. In order to clarify the difference between these two terms, I refer to reference¹ of L. Valdivielso). With regard to the “Estimation of probabilities”, it should be noted that once the model is decided upon, the available information should be used with objectivity when the probabilities are assigned, with care being taken not to confuse reality with desire, nor the probability of the event with the utility of the outcome. As regards the “Assignment of utilities”, it should be noted that unlike the assignment of probabilities, when utilities are assigned, the data provided by observation or external guidelines should be disregarded and one should think for oneself. Utility is a parameter unique to each person and his or her personal universe. The replacement of utility by other more reproducible, “objectifiable” parameters alien to the individual factor does not affect the demands of restricted rationality, but it is unacceptable when we try to address problems with wider rationality.

In addition, the example of a “standard lottery” is proposed which, although based on the so-called theory of expected utility,² is “a difficult technique for the subjects surveyed, and values could be contaminated by the risk aversion of the surveyed subjects”.³ The term “Russian roulette” is inappropriate in our view. We therefore think that the values of both probabilities and utilities assigned in our article are perfectly justified.

As regards the results, although it is true that the 8.8–9.8% decrease in women who have a miscarriage could be considered adequate in relative terms, we think that the absolute term approximation, resulting in a 0.4–0.46% reduction, is much more adequate for this study. In addition, the observation regarding universal screening has not been omitted, but is implicit in the data on preterm deliveries and the absolute reductions associated with universal screening.

Finally, we should like to state that the L. Valdivielso's conviction that the benefits of TSH measurement do not require a formal cost-effectiveness analysis goes against the principles of evidence-based medicine.⁴