March 2018: Use of salivary cortisol and cortisone in the high and low dose synacthen test

Rob Harrison 200x269

By *Charlotte J Elder, *Robert F Harrison, Alexandra S Cross, Ruben Vilela, Brian G Keevil, Neil P Wright, Richard J Ross
*Authors contributed equally
Clinical Endocrinology
DOI: 10.1111/cen.13509

This project is about sticking needles into infants and children or, more precisely, about not doing so. Cortisol is an essential stress hormone and its deficiency without treatment results in death through adrenal crisis. Children can be born with adrenal insufficiency (congenital AI) or it can be acquired in both children and adults, where it is associated with weight loss, muscle weakness, fatigue, low blood pressure and, ultimately, death through adrenal crisis. Early diagnosis and treatment in children is vital.

Checking whether or not symptomatic people have enough serum cortisol (cortisol in the blood) is the primary means of diagnosis and this is traditionally done by introducing a bolus (“shot”) of a cortisol stimulant and measuring serum cortisol levels before and after – in the world of systems and control what, roughly speaking, we’d call an impulse response test. The difference in level indicates to the clinician if the patient is producing enough cortisol.

The difficulty is that the stimulation and measurement (actuation and sensing) requires venepuncture – a needle into a vein, either one for each stage or both via a cannula – either way: Ouch! This is done around one hundred million times annually, world-wide across all age groups, and there are risks. NHSBT figures suggest that one in three thousand are at risk of nerve damage in the highly controlled arena of blood donation, so, while risks are low, the numbers add up. In addition, venepuncture in children is more difficult still and in infants can be very challenging, requiring specialized techniques and needles. And anyway, who wants to stick needles into kids?

So, what if it were possible to remove the needle from the whole business? Would it be possible to introduce the stimulant through a nasal spray, say, and to observe the result in a saliva sample to achieve the same outcome? – a “sniff & spit” test. This paper focuses on the “spit” side, using data modelling to infer the link between serum cortisol levels and salivary levels of a related hormone, cortisone. A particular difficulty in this work is the dynamic nature of the data – usually such inferential models rely on patients being in a steady-state but, here, it has to work during a highly dynamic phase. Work on the “sniff” side is the subject of a larger study of which this forms part.

By using an appropriate model building strategy we have derived what turns out to be a remarkably simple model (always the best sort) that, according to our analysis, does the trick. The aim of the study was to establish the inferential measurement of serum cortisol as a first step towards “sniff & spit”. In addition we have been able to suggest a revised time for making the most useful observation and have revealed knowledge of the effectiveness of different dose-levels – after all, it is always best to administer the least amount that works.

Of course, more experiments are needed to confirm the use of “sniff & spit” in the field but mighty oaks from little acorns grow. In case you are worried, no children were harmed in this study, healthy adult volunteers gave up their time to help – we thank them.

The potential value of this approach to the problem of testing for AI has already been recognized through a published commentary in the same journal.