Salivary steroid assays: still awaiting a killer application

In this issue, ¹ Peter Wood addresses the issue of the current status of assays for steroids in saliva. I must declare an interest, having been associated with two previous reviews of this topic in 1984 ² and 1993. ³ Katz and Shannon ⁴ heroically demonstrated that corticosteroids could be estimated in saliva in the 1960s, but given the poor sensitivity of the analytical technique – mandating impractically large sample sizes – their pioneering studies passed little noticed. The introduction of radioimmunoassay techniques for steroid estimation made salivary assays more practical, one of the first publications coming from Landman et al. in 1976. ⁵ The Tenovus Workshop on this topic held in 1982 ² resulted in dozens of publications, the majority concerned with cortisol. Certain criteria were also emerging:

Is the assay chemically valid with respect to sensitivity, precision and accuracy?

Do salivary concentrations correlate with plasma or plasma-free concentrations?

Are the results biologically plausible or clinically helpful?

Since then, particularly for salivary cortisol determinations in the context of screening for Cushing's syndrome, much effort has gone into the following criterion:

What is the clinical sensitivity and specificity of the result?

Wood has proposed certain criteria for diagnostic salivary steroid testing, which include ‘Single saliva samples should be just as informative as single serum samples’. I would totally reject this criterion since it negates the main advantage of salivary sampling, namely the ethical and practical possibility of collecting multiple samples. For instance, it has proved practical to collect daily saliva samples from women over about 30 days and to determine progesterone concentrations to plot a profile. While ‘a single salivary progesterone estimation is not as informative as a single serum progesterone measurement’, ¹ this is countered by the fact that a profile is far more informative than a single estimation – no matter how accurate the latter is, particularly as many women have irregular cycles. Similarly, in a patient with 21-hydroxylase deficiency, a diurnal profile of four to six salivary estimations of 17-hydroxyprogesterone is more informative than a single plasma estimation. ⁶

Wood has identified 10 published studies confirming that evening salivary cortisol concentration can be used as a reliable screen for Cushing's syndrome. Since there are many more earlier, perhaps anecdotal, studies suggesting the same thing dating back over 30 years, and the undoubted ethical and economic advantages of non-invasive sampling, the following question must arise: ‘Why is the salivary cortisol assay not in routine use in the majority of district general hospitals?’ One answer is perhaps that laboratories assay the samples that they receive for the analytes specified on the forms. The initial screening for Cushing's syndrome may be performed by general practitioners or general physicians, each of whom will see relatively few such patients. They will see little advantage in changing their practice and will continue to send blood samples. Another reason is that laboratories have increasingly moved to automated analysers to cope with increased workload and reoptimization of staff profile – otherwise known as deskilling. There is a reluctance to introduce a manual assay that will tie up a skilled technician for an afternoon. If suitable commercial reagents are introduced for the major analysers, the position may change. A letter in this edition of the Journal ⁷ describes an assay for salivary cortisol on the Abbott Architect analyser, but does not state explicitly that it will be generally available. The problem appears to be that companies are waiting for the killer application, a salivary assay that everyone will need, before committing to development and marketing.

Given that Wood asks the question, research or routine?, and that even salivary cortisol – the best validated salivary steroid analyte – has failed to achieve ‘routine’ status, I would answer ‘research’ and briefly consider the validity, biological plausibility and clinical value of assays for other steroids.

If concentrations in saliva appear to not reflect plasma total or plasma-free concentrations, the problems may lie in two distinct areas. Firstly, current assays, even when reoptimized for adequate sensitivity, may lack the specificity required for accurate determination. Secondly, even using accurate assays, the correlation may be poor due to blood contamination or metabolism, particularly metabolism to the analyte.

Assays for salivary 17-hydroxyprogesterone and androstenedione appear to be valid and to give valuable information for optimization of therapy. They are in routine use in the small number of centres providing clinical advice to patients with the rare conditions of congenital adrenal hyperplasia.

If the clinical biochemistry laboratory is to contribute to assessment of fetoplacental function, then the regular estimation of unconjugated oestriol in saliva is probably the best way to do it. However, in Europe, changes in radiological and obstetric practice have made the method obsolete.

For salivary oestradiol and testosterone in female subjects, I would argue that since we do not have analytically valid plasma assays at this time, it is unrealistic to expect valid assays for salivary concentrations. This is not entirely a matter of sensitivity: assays for salivary testosterone in male adolescents generate results that are biologically plausible, e.g. they correlate with Tanner ratings. ⁸ As for aldosterone, few laboratories determine this steroid in plasma, and the average centre will see few new patients each year in whom it is necessary to refer samples. In such patients, venesection for determination of electrolytes and renin is usually required, negating any advantage of saliva as a non-invasive technique.

For salivary progesterone, Wood has presented arguments against using salivary samples taken on one day, assayed singly, in multiple samples or pooled, to check for ovulation or assess luteal function. Thus there is little role for such determinations in routine clinical practice. However, in research studies of the effects of nutrition and exercise on ovarian function and of the onset of menarche, the collection and assay of salivary samples for 30 days, or menses to menses, and the plotting of a profile, is valuable. Even if a commercial automated assay were available, however, the cost of running monthly profiles in infertile patients to provide a routine service would probably be prohibitive.

I find little value in salivary dehydroepiandrosterone (DHEA) assays. The major use for the plasma dehydroepiandrosterone sulphate (DHEAS) assay is to characterize virilization in female subjects, particularly to screen for carcinoma. There seems little justification to introduce another variable into this critical test, particularly as the correlation between the plasma and saliva concentrations of a polar unconjugated steroid such as DHEAS would be expected to be poor. Indeed my own study ⁹ of concentrations in individuals receiving a DHEAS infusion – to achieve a broad concentration range and avoid oral contamination – showed a poor correlation, not only between blood and salivary concentrations of DHEAS but also between whole saliva and parotid fluid concentrations, the latter being collected using a Curby cup. The estimation of unconjugated DHEA is fraught with difficulty, given the much higher concentration of DHEAS and the ubiquity of sulphatases.

In conclusion, no assays for salivary steroids are in general routine use. Barriers to the adoption of the salivary cortisol assay are logistical and psychological, not scientific. Assays for androstenedione and 17-hydroxyprogesterone are valid but quite properly confined to those centres with a significant sample throughput and experience in interpreting the results. There is a role for salivary progesterone in field studies, long-term studies in athletes and studies in children but probably not in a routine infertility service. Routine assays for other steroids must await a proven clinical need, superior assays and superior publications.