Article Text
PDF
Abstract
Objective The short synacthen test (SST) is widely used across the UK to assess adrenal reserve but there remains no consensus on the timing of cortisol sampling to help diagnose adrenal insufficiency. The main objective of our study was to see if both 30 and 60 min sample are required following administration of synacthen to investigate suspected adrenal insufficiency (AI).
Design This was a single-centre retrospective study of 393 SSTs measuring 0, 30 and 60 min cortisol levels after administration of 250 µg of synacthen.
Patients and methods All the SSTs for patients suspected of primary or secondary AI between April 2016 and October 2018 were included in this study. The tests were performed as per our hospital protocol. A post-adrenocorticotropic hormone (ACTH) cortisol response of 420 nmol/L at any time point was considered adequate to rule out AI. The data were analysed to ascertain the proportion of patients who achieved this level at 30 and/or 60 min.
Results A total of 393 SST results were included in this study. Patients were divided into two groups depending on whether (group A) or not (group B) they were on steroids. Overall, a total of 313 (79.6%) subjects achieved cortisol level of ≥420 nmol/L at 30 and 60 min while 19 (4.8%) had late response (ie, insufficient 30 min cortisol levels, rising to ≥420 nmol/L at 60 min). Another 61 subjects (15.5%) showed insufficient response at both 30 and 60 min (ie, failed to achieved level of ≥420 nmol/L). Importantly, there was no patient in either group who had adequate response at 30 min and then failed at 60 min. Patients in group A were more likely to have inadequate response at both 30 and 60 min while patients in group B were more likely to have normal response at both time points.
Conclusions Our results suggest that about 5% of people undergoing SST may be inappropriately diagnosed as having AI (and subjected to long-term unnecessary steroid treatment) if the 60 min sample is not maintained. We suggest that 30 min sample does not add any additional diagnostic utility and can be omitted thus simplifying SST even further and saving on cost and resources. We propose that single measurement after 60 min of administration of synthetic ACTH is a sufficient screening test for AI.
- synacthem
- adrenal
- ACTH
- pituitary
Statistics from Altmetric.com
Introduction
The diagnosis of adrenal insufficiency (AI) remains challenging and is usually based on the finding of a very low morning serum cortisol or inappropriate responses to adrenal stimulation with synthetic ACTH, called synacthen, in the right clinical setting.1 2
The short synacthen test (SST), also called cosyntropin or short tetracosactide test, was introduced >50 years ago for the assessment of suspected primary AI.3 The test utility has grown considerably over the years and it is now also used to assess secondary AI.4 The insulin tolerance test (ITT) remains the ‘gold standard’ test to assess integrity of the hypothalamic–pituitary–adrenal (HPA) axis. However, this test is too labour intensive, relatively risky, has several contraindications and is performed only very infrequently.
There are two version of SST—high-dose SST involves administration of supraphysiological 250 µg synacthen as opposed to low-dose SST in which 1 µg synacthen is administered which is considered more physiological. Both of these tests are used in clinical practice and have similar diagnostic accuracy.5 The high-dose SST is used very widely to assess HPA function and its use has risen from 24% in 1988 to 50% in 19946 and to 98% in 2018.7 It is now considered as the ‘standard’ for the diagnosis of AI2 due to its simplicity.2 The commercially available synacthen is available in 250 µg/mL ampoules making the high-dose test easy to perform while low-dose test requires large complex dilutions methodology that is prone to dosing miscalculation and is subject to potential errors through loss of material due to poor intravenous technique resulting in incomplete injection.8
The SST is often carried out by the nursing staff thus saving on doctors’ time. It is well tolerated by majority of patients, is performed in an outpatient setting and involves administration of an intramuscular or intravenous injection of synacthen and collecting blood sample for serum cortisol at 0, 30 and (sometimes) 60 min after administration. Importantly, there remains considerable variation in the methodology especially with regards to sample timings with some units taking sample for serum cortisol at 0 and 30 min while other units take sample at 0, 30 and 60 min. First national UK audit of the SST published in 2010 demonstrated this variation in practice across the UK.9 There is an obvious risk of erroneous diagnoses of AI due to this inconsistency in sampling times for the SST.10 It is pertinent to note that the only time point that has been validated against the ITT is 30 min sample11 which has led many to believe that 60 min sample should not be used to assess AI.12 In an analysis of 767 SSTs from a university hospital, authors concluded that 60 min sample was not required in 96% cases.13 However, there are others who report that 60 min value is a better predictor of adrenal function.14
The main objective of our study was to assess the utility of 30 and 60 min sample in the SST. In particular, we set out to establish how often a patient would be misdiagnosed as being normal or having AI if the 30 or 60 min sample was omitted.
Subjects and methods
Subjects
A retrospective single-centre study conducted at Bedford Hospital NHS Trust, Bedford, UK. After careful consideration, we decided not to seek patient consent or ethical approval as the data collection was retrospective and all data were completely anonymised. The study was undertaken as a quality improvement project and all tests were carried out in accordance with the local protocol and informed consent was obtained for each test.
All subjects who had the SST conducted to assess adrenal reserve between April 2016 and October 2018 were included in the study. Patients’ records and basal and post-ACTH cortisol values were extracted from the electronic database held in the hospital.
Patients who were on steroid which had not been omitted prior to the test, those taking HRT or combined contraceptive pill/oral contraceptives (OCP), pregnant women, children under 16 years of age and those who had an incomplete test, that is, where serum cortisol level was not measured/available at all three time points (0, 30 and 60 min) were excluded from the study.
The SST protocol
There is a written protocol in our trust for SST which is followed across the hospital and involves measuring serum total cortisol (nmol/L) at 0 min. A supraphysiological dose of 250 µg synacthen (Synacthen; Mallinckrodt Speciality Pharmaceuticals Ireland) is then administered intravenously or intramuscularly and further blood samples are taken at 30 and 60 min.
The tests are performed in the morning between 09:00 and 11:00. Although we do not ask patients to fast overnight but they are advised to refrain from eating, drinking or smoking for the 30 min before the test. They are rested in a sitting position for 15 min before the test and then for the total duration of the test. Premenopausal women are usually tested within first 7 days of their menstrual cycle unless there is an urgent indication to perform the test. In case of an ongoing steroid replacement therapy, all patients are switched to hydrocortisone at least 2 months prior to the SST which is stopped at least 12 hours before testing.
The SST has a major limitation in that it cannot detect acute pituitary failure.15 Our hospital is not a neurosurgical centre and we have no patient who requires assessment of HPA axis in an acute postpituitary surgery situation. In our study, there was no patient who had received pituitary radiotherapy in the preceding 2 years.
Analytical methods
There are number of assays now available to measure serum cortisol and it is now well established that differences in the assays does have a major impact on the interpretation of cortisol values in SST.16 In the early days, when the SST was first described in 1960s, a non-specific, fluorimetric assay was used that measured not only cortisol but corticosterone also resulting in significantly higher cortisol results.17 However, modern cortisol immunoassays are fully automated and each has its own performance characteristics and specificity for cortisol and may provide results that will vary considerably depending on detection antibodies. A peak cortisol level of 550 nmol/L after ACTH administration is usually considered as sufficient evidence for normal adrenal function, although it may range from 418 to 574 nmol/L depending on the assay used.18 There may also be other factors that may influence the interpretation of serum cortisol (eg, use of OCPs, HRT, nephrotic syndrome, liver disease and so on). It is therefore very important to have assay-specific serum cortisol cut-offs to ensure accurate interpretation of SST.2 19
Although the cortisol cut-off used in SST are derived from non-standardised immunoassays a recent study utilising liquid chromatography tandem mass spectrometry that offers direct measurement of steroids has provided lower cut-off levels for cortisol after ACTH stimulation, possibly due to elimination of cross-reactivity between steroid intermediates and cortisol, which in turn reduces the likelihood of false-positive tests for AI.20
In our hospital, serum cortisol is assayed by electrochemiluminescence immunoassay ‘ECLIA’ using a Cobas 8000 analyser (Roche Diagnostics, Mannheim, Germany). The method has been standardised against the Institute for Reference Materials and Measurements/IFCC 451 panel (ID GC/MS, isotope dilution-gas chromatography/mass spectrometry).21
An adequate response to short synacthen is considered if cortisol at 30 or 60 min is ≥420 nmol/L.18 21 The assay has a measuring range between 1.5 and 1750 nmol/L. The reference range for 09:00 cortisol is 135–550 nmol/L, a higher normal cortisol concentration is expected in women who are pregnant or on OCP.
Blood samples are collected in serum separator tube, are allowed to clot and following separation one aliquot of separated serum are stored at 4°C and analysed within sample stability limits. Three internal quality control (IQC) samples (mean target concentrations 122, 464 and 729 nmol/L, respectively) are tested precortisol and postcortisol analysis. All IQC results are persistently within 2 SD of the mean. Cortisol assay has an average uncertainty of measurement of 5.23% and coefficients of variation ratio of 0.62 (desirable limit <1.5).
Our laboratory is enrolled in the UK National External Quality Assurance Scheme and our cortisol assay has consistently performed well within the recommended range. Our laboratory is fully accredited by United Kingdom Accreditation Service and meets ISO 15189 standard for medical laboratories.
Statistical methods
All analysis was completed using the statistical software R25. Packages used were: ggplot2 V.3.2.0, caret V.6.0.84, epiR V.1.0.2, ggpubr V.0.2.1, dplyr V.0.8.1, nlme V.3.1.139 and psych V.1.8.12. Due to a large range of cortisol values measured, values are given as the median with upper and lower quartiles. This prevents singular extreme cortisol values from influencing summary statistics.
Results
A total of 393 individuals with suspected AI who had SST performed (238 female and 155 male patients) were included in the analysis. Six patients were excluded from the analysis due to incomplete data collection. The median age was 60 (IQR=44, 74), the sample was made up of 60.5% female and 40.5% male.
The study population was divided into two categories—those who were on steroids (group A, 120 patients, 30.5%) and those who were not on steroids (group B, 273 patients, 69.5%). The main reasons for pre-existing steroid therapy were either short-term coverage due to a strong clinical suspicion of AI or long-term treatment due to an underlying inflammatory disease (eg, polymyalgia, inflammatory bowel disease, rheumatoid disorders and so on).
We classified response to synacthen in three categories:
An insufficient response if serum total cortisol concentration of ≥420 nmol/L not achieved at 30 or 60 min.
A late response if serum total cortisol concentration of ≥420 nmol/L achieved at 60 min, but not at 30 min.
A normal response if serum total cortisol concentration of ≥420 nmol/L achieved at 30 and/or 60 min.
Table 1 shows the median serum cortisol with IQR at baseline (ie, 0 min, just before the administration of synacthen), 30 and 60 min. Figure 1 depicts boxplots of each of the cortisol measurements at three time points. Figure 2 depicts cortisol levels of all the patients included in this study at three time points (baseline, 30 and 60 min).
The median serum cortisol value with lower and upper quartiles in brackets of cortisol levels at baseline, 30 and 60 min, for group A (patients on steroids), group B (patients not on steroids) and the whole sample
Boxplots for each of the cortisol measurements at baseline (top), 30 min (middle) and 60 min (bottom), split by group A (left), group B (middle) and the whole sample (right). Within each graph, insufficient response (IR) indicates the group with insufficient response at both 30 and 60 min. Late response (LR) indicates the group with insufficient response at 30 min and a normal response at 60 min. Normal response (NR) indicates the group with a normal response at 30 and/or 60 min.
The median cortisol level (black line) for group A (left) group B (middle) and whole sample (right), split by outcome group, for the baseline, 30 and 60 min measurement. Across each graph, insufficient response (IR) indicates the group with insufficient response at both 30 and 60 min. Late response (LR) indicates the group with insufficient response at 30 min and a normal response at 60 min. Normal response (NR) indicates the group with a normal response at 30 and/or 60 min. The shaded region indicates the region of IQR, 50% of the measurements fall within this range. Note the larger spread of data for the ‘normal’ response category. The cut-off value of 420 nmol/L restricts the variation for insufficient and lateresponses.
The outcome of the SST for the group A, group B and the whole sample is plotted in table 2. Each table plots the outcome of the measurement at 30 min against the outcome at 60 min to identify the contingency between the two measurements. There were no observed cases where a patient had a normal measurement at the 30 min measurement with an insufficient adrenal measurement at 60 min.
Comparison of cortisol responses to ACTH at 30 and 60 min. These data show that there is a proportion of patients who will be diagnosed as AI if only 30 min sample was used
Table 3 shows summary statistics (sensitivity, specificity, positive predictive value and negative predictive value) of the contingency tables for each group. Broadly speaking the diagnostic utility of the 30 min measurement performed well against the 60 min measurement. Across both groups, sensitivity of the 30 min measurement was at 1, indicating that all those identified with AI at 60 min were also AI at 30 min. Similarly, specificity was high across both groups. This means that those identified as having ‘normal’ measurements at 60 min are highly likely to have received a normal measurement at 30 min.
Contingency table analysis. Sensitivity is the proportion of patients with insufficient responses at 30 min from those with insufficient responses at 60 min. Specificity is the proportion of patients with normal responses at 30 min from those with normal responses at 60 min. Positive predictive value is the proportion of patients with insufficient responses at 60 min who also have insufficient responses at 30 min. Negative predictive value is the proportion of participants with normal responses at 60 min from those with normal responses at 30 min
The positive predictive value (ie, the probability of an insufficient adrenal response at 60 min given an insufficient adrenal response at 30 min) was particularly low in group B. There were 21 cases identified with AI at 30 min in the group B, 11 of these 21 cases had normal measurements at 60 min. This suggests that the measurement at 30 min does not add any additional information when compared with the 60 min measurement. Note only a small number of patients in group B had an insufficient adrenal response at 30 min.
We used McNemar’s tests to identify significant differences between the measurement at 30 and 60 min for the group A, group B and whole sample separately. For group A, we found a significant difference (p=0.013) between the measurement at 30 and 60 min. The eight patients identified with AI at 30 min with a normal response at 60 min drove this relationship. For group B, we found a significant difference (p=0.002) between the measurement at 30 and 60 min. Again, the 11 patients identified with AI at 30 min with a normal response at 60 min drove the significant difference. For the whole sample, we found a significant difference (p=0.00003) between the measurement at 30 and 60 min. The 19 patients identified with AI at 30 min with a normal response at 60 min drove the significant difference.
Discussion
In our study, there were no cases where a patient had an adequate cortisol response to ACTH at 30 min with an insufficient response at 60 min thus suggesting that 30 min sample may not be necessary. This was true of both group A (patients on steroids) and group B (patients not on steroids). In simple terms, no individual was found to achieve an adequate cortisol response at 30 min and fail to maintain adequate response at 60 min. Conversely, there were 19 patients (about 5%) who were late responders (ie, insufficient response at 30 min but normal at 60 min)—about 7% in group A and 4% in group B.
These findings suggest that a 30 min postsynacthen cortisol measurement does not appear to offer additional diagnostic value in identifying AI. Similar conclusions were drawn in a retrospective study performed in South Asian patients11 and in Caucasian individuals.22 In the later study, Chitale et al concluded, based on retrospective analysis of 384 SST samples, that a significant proportion of patients risk being inappropriately diagnosed as having AI if diagnosis is based on 30 min sample only. In their study also no individual was found to achieve an adequate cortisol response in the 30 min sampling and fail to sustain this response at 60 min.
SST is a simple and safe test and is not resource intensive. However, if we are able to simplify the protocol further to have only one cortisol measurement 60 min after ACTH stimulation, we can reduce labour and costs even further. There have also been attempts to identify basal cortisol level that will predict AI in SST so as to obviate the need to do SST altogether.23
Based on the results of present study, it is not safe to restricting the test to a single 30 min value because it would run the risk of misclassifying a significant number of patients (5%) as having AI who would then be subjected to unnecessary long-term steroid replacement therapy. As far as we are aware there is no study to date to suggest that subject who exhibit a delayed response to synacthen (late responders) benefit from steroid replacement therapy. In a double-blind, placebo-controlled study to assess maximal cortisol secretion rates during cosyntropin (250 µg) stimulation in healthy volunteers, it was found that cortisol levels rise relatively rapidly in the first 30 min and continues to rise for up to 60 min after ACTH administration before reaching a plateau.24
ITT remains the gold standard in the diagnosis of AI. However, the test is very labour intensive and time-consuming with many contraindications making it a much less favoured tests among endocrinologists and SST has largely replaced ITT as the method of choice to investigate suspected for AI except in acute postpituitary failure situations. Recent Endocrine Society Clinical Practice Guideline recommend a peak cortisol levels <500 nmol/L (18 µg/dL) (assay dependent) at 30 or 60 min to indicate AI.2
It is important to note that our study population was a mixed one which encompassed patients with suspected primary and secondary AI. However, we had no patient in the immediate postacute pituitary failure stage. Our patient population does reflect normal clinical practice in hospitals across the UK.
There are some obvious limitations to the present study. Our data collection was based on reviewing patients’ notes and was therefore retrospective. Our aim was not to compare SST with ITT but solely to assess if two cortisol levels post ACTH were really required and offer diagnostic value. It would be desirable to validate our findings against a gold standard test (such as ITT). Our patient population was mixed and heterogeneous (indicative of patient population in a general hospital). However, the study was single-centred and the samples was handled and analysed in an identical manner throughout the duration of study which improves the applicability of results.
Thus although we appreciate that 30 min sample is the one that has been shown to have best correlation with the ITT and many centres across the UK have chosen to retain 30 min sample but based on the results of present study we propose that 30 min sample does not offer any diagnostic utility in the interpretation of SST and should perhaps be abandoned.
Main messages
The 30 min sample in a short synacthen test (SST) can be omitted as it does not add any additional diagnostic value in identifying adrenal insufficiency (AI) regardless of whether patient is previously on steroids or not.
SST protocols which omit the 60 min sample could lead to a false diagnosis of AI in 1 in 20 patients.
The results of this study have the potential to enable further simplification of SST, thus saving on cost and human resources.
Current research questions
It would be useful to establish a cut-off basal serum cortisol level that will exclude the need for a SST to be carried out.
The 60 min sample for the SST should be validated against the gold standard insulin tolerance test.
A similar study should be carried out on a paediatric population.
What is already known on the subject
Short Synacthen Test is widey used for the assessment of Adrenal Insufficiency.
There is lack of consensus on the timing of cortisol sampling.
Acknowledgments
We are grateful to our endocrine nurse (Ann Connelly) and other nurses in our Day Unit who performed the tests as well as our patients. We are also grateful to our IT Lead (Mr Andrew Cakebread) who retrieved data from our Hospital Computer.
References
Footnotes
Contributors RK contributed to the conception and design of the study, and wrote the first draft. PC performed and wrote the statistical analysis. KM collected the data and carried out initial data analysis. ZR contributed to clinical data collection. LW contributed to the conception and design of the study. WW contributed to the conception and design of the study, and the writing of the article. All authors have given their approval for the final version of the manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval After careful consideration, we decided not to seek patient consent or ethical approval as the data collection was retrospective and all data were completely anonymised.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Data are available upon reasonable request.