Abstract
Background
Since more than 90% of cortisol is bound to protein, serum free cortisol (SFC) may be a more appropriate marker of adrenal status than total cortisol. However, measurement of SFC is technically difficult and calculated SFC may offer a more practical alternative.
Methods
SFC, measured by equilibrium dialysis coupled with immunoassay, and calculated using Coolens' equation from total cortisol and corticosteroid binding globulin (CBG) concentrations, was compared in short Synacthen test (SST) serum from 42 patients, of whom 20 demonstrated a suppressed adrenal response.
Results
Considering the patient group as a whole, calculated SFC was found to be significantly lower than measured SFC, pre- and post-Synacthen (P < 0.05 and <0.001, respectively). Upon classifying the patients as pass or fail based on total cortisol response to Synacthen, the difference in calculated and measured SFC only reached statistical significance for post-Synacthen concentrations in the pass group (P < 0.01), suggesting a greater discrepancy at higher cortisol concentrations. There was no difference in CBG levels between the pass and fail groups and both measured and calculated SFC gave a diminished 30 min response in subjects deemed to have failed the SST.
Conclusion
Coolens' equation was found to underestimate measured SFC in this cohort of outpatients, as has been previously demonstrated, particularly in patients with a pronounced acute phase response. Although calculated SFC gave a diminished response in individuals deemed to have failed the SST, the concentration-dependent nature of the discrepancy may limit the usefulness of this method for assessing adrenal status.
Introduction
In plasma, more than 90% of circulating cortisol is bound to protein; approximately 70% to corticosteroid binding globulin (CBG) with high affinity; and the remainder to albumin with low affinity. 1,2 Although the free (non-protein-bound) component is generally accepted to be physiologically active, 1,3 it is still usual to measure total cortisol (free plus bound) when assessing adrenal function.
However, the limitations of total cortisol in this respect are recognized, and interest is growing in serum free cortisol (SFC) as a more appropriate marker of adrenal status, particularly where concentrations of CBG are abnormal. For example, in a number of studies, measurement of SFC has identified a subgroup of patients with sepsis that demonstrates a suboptimal total cortisol response to the short Synacthen test (SST), but an SFC response comparable with normal controls. 4–6
Serum or plasma free cortisol has been measured in recent reports either by equilibrium dialysis, 4,7–10 ultrafiltration, 5,10,11 coupled with immunoassay, 4,7,10 mass spectrometry 8,9 or ligand binding. 5,11 These methods are technically difficult and a simpler alternative to measuring SFC is to calculate an estimate from total cortisol and CBG concentrations using Coolens' equation. 1
A study was recently carried out in our institution to assess whether inhaled corticosteroids (ICS), used in the treatment of lung disease, have a suppressive effect on the hypothalamic-pituitary-adrenal axis. 12 SSTs were performed on 50 outpatients with bronchiectasis (33 on ICS treatment), where 20 patients (16 on ICS) demonstrated an inadequate adrenal response (total cortisol <550 nmol/L, 30 min post-Synacthen), indicating that ICS may be systemically absorbed in some individuals.
Here we report an evaluation of SFC, measured by equilibrium dialysis coupled with immunoassay, and calculated using Coolens' equation, in this same group of patients, including those identified as adrenally suppressed on the basis of the standard SST. Results were compared in the patients as a whole and classified according to adequacy of response to Synacthen (pass/fail).
Methods
SFC and CBG concentrations were measured in serum collected during SST (−25°C for 2 y) from 43 of the patients in the aforementioned study, of whom 20 demonstrated a suppressed total cortisol response to the SST. One patient with a normal SST response had a CBG concentration greater than three standard deviations (SD) from the group mean and was excluded from further data analysis. The patient cohort thus consisted of 31 women and 12 men, aged 64 (10) y (mean [SD]), of whom 15 (10 women, 5 men) had a suppressed adrenal response (suppressed patients aged 64 [12] y).
The SST was performed according to standard clinical protocols; 250 µg of Synacthen (synthetic adrenocorticotropic hormone1–24, Alliance Pharma plc, Wiltshire, UK) was given intramuscularly and samples were collected for serum cortisol measurement immediately prior to, and 30 min following, Synacthen administration. The data were stratified into pass and fail groups based on the total cortisol concentrations measured on the Advia Centaur (Siemens Medical Solutions Diagnostics, Berkshire, UK) in the original study (normal response defined as post-Synacthen cortisol greater than 550 nmol/L). 12,13 The study was approved by South Birmingham Research Ethics Committee.
Serum free cortisol
Free cortisol was separated from bound by equilibrium dialysis. The dialysis devices, constructed following a published method, 10 allowed 1 mL of serum to be dialysed against 0.25 mL of Roche Universal Diluent (Roche Diagnostics Ltd, West Sussex, UK) and were incubated at 37 ± 0.5°C for 21 ± 0.2 h. Cortisol in the dialysable material was assayed using the Roche Modular Analytics E170 platform (Roche Diagnostics). The between-batch coefficients of variation (CVs) for SFC measurement were 4.1% at 111 and 829 nmol/L (n = 7).
The between-batch CVs for the E170 serum cortisol assay were measured as 4.0–7.3% over the concentration range 35–98 nmol/L. When determined for serum diluted in recommended Roche Universal Diluent (2 vol in 5), the CVs were 1.5–5.1% over the range 24–86 nmol/L. The manufacturer's quoted functional sensitivity was 2.0 nmol/L. A lower limit of reporting of 20 nmol/L was used. Serum was serially diluted in Roche Universal Diluent up to a dilution of 1 vol in 32 (28 nmol/L). The mean measured cortisol corrected for dilution and expressed as a proportion of the undiluted value was 101% (range 95–109%).
Corticosteroid binding globulin
Serum CBG concentrations were measured by radioimmunoassay (Biosource Europe, Nivelles, Belgium). At 27 and 48 mg/L of CBG, within-batch CVs were 8.5% and 7.6%, respectively, while between-batch CVs were 13% and 14%, respectively.
Calculated SFC
SFC was calculated from total cortisol and CBG concentrations using Coolens' equation. 1 Serum total cortisol concentrations used in the equation were obtained by re-assay of the samples on the Roche Modular Analytics E170 platform to allow comparability with measured SFC concentrations.
Coolens' equation is based on two simultaneous cortisol–protein binding equilibria: binding to CBG that can be saturated and binding to albumin that cannot. The equation accounts for both the concentration of CBG and saturation of cortisol binding (at approximately 650 nmol/L), assuming an affinity constant of K = 3 × 107 /mol/L at 37°C, with results expressed in molar terms. The ratio of free cortisol bound to albumin is assumed to be constant at N = 1.74.
1
A simplified version of the equation was used in this work, where U, G and T represent SFC, CBG and total cortisol, respectively, with all concentrations expressed in µmol/L:
1
Statistical methods
All statistical analyses were performed using Analyse-it (Analyse It, Leeds, UK). The Shapiro-Wilk W test was used to determine whether data were normally distributed. Parametric data are presented as mean (SD) and non-parametric data as median (25th–75th percentile). In the statistical analyses, SFC concentrations of <20 nmol/L were included as equal to 20 nmol/L. When comparing means from two groups of parametric data, the unpaired t-test was employed. Groups of non-parametric data were compared using Mann-Whitney U analysis. Graphical data were fitted using linear regression. A P value <0.05 was interpreted as statistically significant.
Results
Total cortisol concentrations
Pre- and post-Synacthen total cortisol concentrations for the entire group, as measured in the original study, 12 were as follows: median (25th–75th percentile) 373 (313–456) nmol/L and 552 (502–626) nmol/L, respectively. The basal and post-Synacthen total cortisol values in the group that passed the SST were 430 (372–502) nmol/L and 619 (568–737) nmol/L, and in the group that failed the SST were 347 (296–373) nmol/L and 496 (463–515) nmol/L. Pre- and post-Synacthen total cortisol concentrations for the entire group, assayed on the Roche E170, were higher: 480 nmol/L (410–536 nmol/L, P < 0.001) and 769 nmol/L (711–830 nmol/L, P < 0.0001), respectively. Roche basal and post-Synacthen total cortisol concentrations in those that passed the SST were 489 (428–619) nmol/L and 819 (758–888) nmol/L, and in those that failed were 466 (367–509) nmol/L and 732 (623–782) nmol/L.
SFC concentrations
All data
Figure 1 shows plots of measured versus calculated SFC for all SST data.
Plots of measured versus calculated SFC for all SST data (n = 42). Diamonds and triangles depict pre- and post-Synacthen data, respectively. (a) All data fitted with linear regression, long dashes depict line of equivalence, solid line the regression and dotted lines the 95% confidence intervals. (b) Separate fits for pre- and post-Synacthen data (shown in black and grey, respectively), where solid lines are the regression and dotted and dashed–dot lines, respectively, represent 95% confidence intervals. SFC, serum free cortisol; SST, short Synacthen test
From Figure 1a, calculated SFC was found to be consistently lower than measured (y = 0.74x + 0.92; 95% confidence intervals: [slope] 0.66–0.82; [intercept −5.2 to 7.0). Separate fits of pre- and post-Synacthen data revealed notably more scatter for the latter (R 2 = 0.77 versus 0.68), which was reflected in wider confidence intervals for the intercept (−4.9 to 6.1 versus −23 to 11) but not for the slope (0.66–0.93 versus 0.63–0.98).
Taking into consideration all data, the difference between measured and calculated SFC concentrations reached statistical significance for both pre- and post-Synacthen (median [25th–75th percentile]: 28 [23–46] nmol/L versus 25 [19–36] nmol/L, P < 0.05 and 90 [71–106] nmol/L versus 59 [42–87] nmol/L, P < 0.001, respectively). However, the discrepancy was noted to be greater at the higher, post-Synacthen SFC concentrations.
Data classified as SST pass or fail
Measured and calculated SFC concentrations were then compared in patients classified as pass or fail on the basis of total cortisol response to Synacthen (Table 1).
Pre- and post-Synacthen cortisol data (analysed on the Roche E170) for the SST pass and fail groups
SFC, serum free cortisol; SST, short Synacthen test
P value refers to measured versus calculated SFC. Non-parametric data described as median (25th–75th percentile)
Stratifying the data in this manner, the difference between calculated and measured SFC only reached statistical significance for post-Synacthen concentrations in the pass group, which was noted to include the highest SFC values.
Using either approach to SFC, the fail group had significantly lower concentrations than the pass group, both pre- (measured and calculated: P < 0.05) and post- (measured: P < 0.001; calculated: P < 0.01) Synacthen. However, a greater degree of overlap between the interquartile ranges of the pass and fail groups was noted for calculated compared with measured post-Synacthen SFC. Accordingly, a greater number of patients that had failed the SST had calculated post-Synacthen SFC concentrations within the absolute range of passes (16/20 within pass range [32–183 nmol/L]) compared with measured SFC (12/20 within pass range [67–183 nmol/L]).
It was noted that only pre-Synacthen total cortisol concentrations were found to have a normal distribution in line with the findings of Clark et al. 13
CBG concentrations
There was no statistically significant difference between CBG levels of the pass and fail groups (mean [SD], pre-Synacthen: 48 [10] mg/L versus 54 [12] mg/L, P = 0.08; post-Synacthen: 48 [11] mg/L versus 52 [15] mg/L, P = 0.36). Furthermore, no difference in pre- versus post-Synacthen CBG concentrations was found across the whole patient group (51 [11] mg/L versus 50 [13] mg/L, P = 0.86), although variation in CBG between the two time points was observed in individual patients. The range of CBG concentrations measured in the group was noted to be wide (28–85 mg/L) when compared with the reference range quoted by the assay manufacturer (31–53 mg/L) (Biosource Europe, Nivelles, Belgium).
Discussion
In this study we have demonstrated a significant correlation between measured SFC and the calculated value based on Coolens' equation in a group of outpatients, receiving ICS therapy. The calculated value of SFC was found to be lower than the measured. However, when taking into account the classification of the patients as those with or without evidence of adrenal suppression, the difference between measured and calculated SFC in the groups was only significant for those with no evidence of adrenal suppression.
Coolens' equation is a simplification of cortisol–protein binding equilbria and assumptions of the equation, e.g. constant binding to albumin, a single CBG–cortisol affinity constant or no consideration of competing ligands, are likely to contribute to the differences between measured and calculated SFC observed. Any discrepancy may be amplified at higher SFC concentrations due to the exponential rise in SFC with increasing total cortisol concentration once CBG is saturated (at approximately 650 nmol/L of cortisol). 14
Such a phenomenon may be further exacerbated post-Synacthen administration, due to the release of other steroid species (e.g. cortisol precursors), which may have differing affinities for CBG/albumin not predicted by Coolens' equation. Moreover, the effect may vary between individuals resulting in the increased scatter observed post-Synacthen in the measured versus calculated SFC plot.
In this patient cohort, the effect of residual inhaled steroid should also be considered since it may have different protein-binding characteristics compared with endogenous cortisol. However, a significant effect was reasoned to be unlikely as the patients had stopped ICS at least 24 h prior to the SST and ICS only have a short half-life in plasma (e.g. beclomethasone dipropionate approximately 3 h). 15 Furthermore, the subjects had not received oral steroids for at least three months prior to the study.
Others have also compared calculated and measured SFC. 5,10 Vogeser et al., 10 in agreement with our findings, reported a calculated SFC that was lower than measured, particularly in patients with a pronounced acute phase response. On the contrary, a group studying patients with sepsis 5 found close agreement between measured and calculated SFC. However, comparison of studies is difficult given the differences in methodology of the ‘free’ hormone assays. The differences in the separation step (equilibrium dialysis, ultrafiltration) and the cortisol assay used (immunoassay, tandem mass spectrometry, ligand binding using tritiated cortisol) may be further complicated by the potentially varying effects of the patient groups studied.
CBG concentrations were found to be wide ranging but, on average, high normal with respect to the assay manufacturer's reference range. Women on the oral contraceptive pill had been excluded from the study in the initial selection process, eliminating this as a potential explanation. It is possible that higher CBG concentrations are associated with long-term steroid use or even relate to chronic inflammation; however, this would be in direct contrast to the profound decrease in CBG in acute inflammation. 4,5,16 Alternatively, they may reflect a characteristic of the study cohort, i.e. predominantly postmenopausal women. Overall no change in CBG concentrations was observed post-Synacthen, which is in contrast to the observations of Lewis et al., 17 who report a small but significant decrease. Clearly, further studies are required to explore this issue further.
Overall pre- and post-Synacthen SFC concentrations were significantly lower in the fail group than those observed in the pass group, demonstrating that generally SFC concentrations reflect total cortisol concentrations in this cohort of patients. This can be explained by the finding that CBG concentrations were not significantly different between the pass and fail groups. However, there was substantial overlap between the pass and fail groups, post-Synacthen, for both measured and calculated SFC, indicating that classification of individual patients would not be identical on the basis of SFC. This is probably due to the wide range of CBG concentrations observed in this patient group. Assay-related differences between the Roche, on which SFC measurements were based, and Centaur measurements, used to classify the data, may also contribute to some extent. Increased overlap was observed for calculated with respect to measured SFC, which is likely to reflect the finding that calculated SFC increasingly underestimates measured SFC at the higher post-Synacthen cortisol concentrations observed in the pass group.
The total cortisol concentrations measured on the Roche for use in Coolens' equation were noted to be markedly higher than those measured on the Centaur in the original study. This highlights the importance of using a method-related cut-off to define a normal response to the SST. 13 Classification of the patients into pass or fail following the SST was thus based on the Centaur measurements, employing a cut-off that has been validated using a robust dataset. 13 The fact that in this work both measured and calculated SFC were based on Roche cortisol values allowed direct comparison of the results.
A potential limitation of this study is that the SST serum had been stored for two years prior to SFC analysis, albeit under identical conditions. However, percentage free cortisol has been shown to be stable for at least one year frozen, 18 and in-house studies have demonstrated that both SFC and CBG concentrations are stable to at least six freeze–thaw cycles.
From a practical aspect, measuring SFC by equilibrium dialysis is time-consuming, labour-intensive and not suitable for high-throughput, routine use. Coolens' equation overcomes many of these problems. However, CBG assays are not widely available and have substantial analytical variation.
Conclusions
A relationship between calculated and measured SFC was confirmed in this cohort of outpatients, although the reason for calculated SFC being lower is yet to be determined. The relationship differed with SFC concentration and potential mechanisms include: limitations of Coolens' equation; effect of inhaled steroids; and lack of specificity of immunoassay. These findings, and indeed the definition of a ‘normal’ total cortisol response to Synacthen, need to be confirmed by measuring cortisol by tandem mass spectrometry and in other patient groups, but may potentially limit the usefulness of Coolens' calculated SFC as a marker of adrenal status.
Both measured and calculated SFC gave a diminished 30 min response in the subjects deemed to have failed the SST based on total cortisol, although the classification was not identical. The clinical significance of this is uncertain and studies relating SFC concentrations to patient outcomes are needed.
DECLARATIONS