The Short Cosyntropin Test Revisited: New Normal Reference Range Using LC-MS/MS

Abstract

Background

The cosyntropin test is used to diagnose adrenal insufficiency (AI) and nonclassical congenital adrenal hyperplasia (NCCAH). Current cutoffs for cortisol and 17-hydroxyprogesterone (17-OHP) are derived from nonstandardized immunoassays. Liquid chromatography tandem mass spectrometry (LC-MS/MS) offers direct measurement of steroids, prompting the need to re-establish normal ranges.

Objective

The goal of this study was to define cutoff values for cortisol and 17-OHP in serum by LC-MS/MS 30 and 60 minutes after intravenous administration of 250 µg tetracosactide acetate to healthy volunteers and to compare the results with LC-MS/MS with routine immunoassays.

Methods

Cosyntropin testing was performed in healthy subjects (n = 138) and in patients referred for evaluation of adrenocortical function (n = 94). Steroids were assayed by LC-MS/MS and compared with two immunoassays used in routine diagnostics (Immulite and Roche platforms). The cutoff level for cortisol was defined as the 2.5% percentile in healthy subjects not using oral estrogens (n = 121) and for 17-OHP as the 97.5% percentile.

Results

Cortisol cutoff levels for LC-MS/MS were 412 and 485 nmol/L at 30 and 60 minutes, respectively. Applying the new cutoffs, 13 of 60 (22%) subjects who had AI according to conventional criteria now had a normal test result. For 17-OHP, the cutoff levels were 8.9 and 9.0 nmol/L at 30 and 60 minutes, respectively.

Conclusions

LC-MS/MS provides cutoff levels for cortisol and 17-OHP after cosyntropin stimulation that are lower than those based on immunoassays, possibly because cross-reactivity between steroid intermediates and cortisol is eliminated. This reduces the number of false-positive tests for AI and false-negative tests for NCCAH.

Adrenal insufficiency (AI) comprises primary (PAI) and secondary adrenal insufficiency (SAI). Diagnosis of AI is based on laboratory tests and can be challenging particularly when insufficiency is partial, as is often the case in SAI. A correct diagnosis is important because AI most often requires lifelong corticosteroid replacement therapy.

The cosyntropin stimulation test has been widely used to diagnose PAI and SAI since the 1960s (1) and is considered the gold standard for the diagnosis of PAI (2, 3). Although it is often used in SAI, the more cumbersome insulin-induced hypoglycemia test is considered the gold standard to diagnose SAI (4, 5). The standard cosyntropin test is performed by administering 250 µg cosyntropin intravenously or intramuscularly. Serum cortisol (s-cortisol) is measured before and 30 and 60 minutes after a bolus injection. Current guidelines define an increase in cortisol above 500 to 550 nmol/L after 30 or 60 minutes as a normal response (3). The cosyntropin test is also used to diagnose partial enzyme defects in nonclassical congenital adrenal hyperplasia (NCCAH). An increase in serum 17-hydroxyprogesterone (s-17-OHP) above 30 to 45 nmol/L 60 minutes after cosyntropin stimulation is considered diagnostic of NCCAH (6).

The current cortisol and 17-hydroxyprogesterone (17-OHP) cutoff levels are based on old, nonstandardized immunological methods that are no longer in use. In recent years, liquid chromatography tandem mass spectrometry (LC-MS/MS) assays for measuring steroid hormones have been increasingly implemented and are now recommended as the gold standard (7, 8). LC-MS/MS offers superior analytical specificity compared with immunoassays and is traceable to one accuracy basis (i.e., international reference materials), enabling comparison of results across methods, time, and location. However, when switching to a new analytical principle, new reference ranges are required, a point recognized and emphasized by the Endocrine Society (9, 10). Because reliable reference ranges and clinical action limits for the cosyntropin test do not exist (11, 12), we aimed to find the optimum cutoff level of s-cortisol and s-17-OHP that would distinguish normal subjects from those with AI and NCCAH, respectively. Furthermore, we directly compared the results of LC-MS/MS with currently used routine immunoassays to evaluate their performance.

Subjects and Methods

Subjects and study design

From June 2016 to August 2017, we included consecutive patients referred for evaluation of AI and NCCAH to the tertiary endocrine specialist center at Haukeland University Hospital, Bergen, Norway. Healthy control subjects were recruited from hospital and university staff. Sixty patients with suspected AI, 34 patients with clinical suspected NCCAH, and 138 healthy control subjects were enrolled (Table 1). Seventeen healthy control subjects were on oral estrogens. Patients using cortisone acetate did not take the medication on the day the test was performed, and those on prednisolone were without medication for at least 72 hours. Patients tested for AI due to long-term use of opioids did not take opioids on the test day.

Table 1.

Characteristics of Patients and Healthy Control Subjects

	Healthy Control Subjects a	Suspected AI	Suspected NCCAH
Patients, n	121	60	34
Women, n (%)	67 (55)	36 (60)	33 (97)
Age, y (range)	40 (23–68)	48 (19–86)	27.5 (17–66)
Systolic blood pressure, mm Hg (range)	128 (109–170)	120 (85–181)	110 (90–125)
Diastolic blood pressure, mm Hg (range)	78 (58–99)	75 (60–105)	74 (60–82)
Smokers, n (%)	5 (4.1)	14 (23)	4 (12)
Body mass index, kg/m2 (range)	23 (15–33)	25 (17–48)	24.5 (20–39)
Creatinine, µmol/L (range)	70 (50–100)	68 (47–244)	62.5 (47–84)

Categorical data are given as n (%); continuous data are given as median (range).

^a

Subjects on oral estrogens excluded.

All study subjects performed a standard short cosyntropin test. Blood samples for s-cortisol and s-17-OHP were taken before (0 minutes) and 30 and 60 minutes after intravenous administration of 250 µg tetracosactide acetate with the patient placed in the recumbent position. The test was performed on nonfasting patients between 8:00 am and 12:00 pm. If abnormal test results were obtained, further diagnostic workup according to European guidelines for AI (3) and NCCAH (6) were performed, depending on the pretest probability for these diseases. The study subjects were categorized as PAI or SAI, NCCAH, or healthy. The test was repeated in 20 healthy persons to assess reproducibility. All samples were immediately centrifuged and stored at −80°C.

Glucocorticoid assay

Samples were assayed for cortisol and 17-OHP by LC-MS/MS as described in detail previously (13). Total analytical variation [coefficient of variation (CV)] for cortisol was <7.4 (97% accuracy), and total CV for 17-OHP was <9.7 (101% accuracy). Our LC-MS/MS laboratory is contributing to an external quality control program (NEQAS for serum steroids). Additionally, the accuracy of our LC-MS/MS assay is regularly verified by analyzing certified reference materials. The accuracy according to the certified reference materials is 105% for both cortisol and 17-OHP.

Samples were also assayed for cortisol by Roche Modular (Roche Diagnostics, Rotkreutz, Switzerland) and Immulite 2000 xpi (Siemens Health Care, Erlangen, Germany) in 54 healthy individuals. Immulite had a total CV of 7% at two concentration levels (280 and 680 nmol/L). Roche Modular had a lower quantification limit of 3 nmol/L and a total CV of 2% at 300 and 700 nmol/L.

Statistical analyses

The subject characteristics are reported as median (range). Because the data were not normally distributed, cutoffs for s-cortisol and s-17-OHP after cosyntropin were obtained by applying nonparametric statistics. The Mann-Whitney U test was used to compare groups. The 2.5th centile of cortisol after 30 and 60 minutes in healthy control subjects not using oral estrogens was defined as the lower normal limit. The 97.5th centile of 17-OHP after 30 and 60 minutes in healthy control subjects was defined as the upper normal limit. Receiver operating curves were also used to calculate the cutoff for 17-OHP. Spearman’s ρ was calculated for the parameters measured with repeated testing in 20 subjects and to determine correlations between LC-MS/MS and the different immunoassays. The reproducibility for s-cortisol and s-17-OHP was also assessed by determination of intraclass correlation coefficient (ICC) [95% confidence interval (CI)] using ln-transformed data and a random-effects mixed model with participant identification as the random variable. An ICC <0.40 is considered as poor reproducibility, 0.40 to 0.75 as fair-to-good reproducibility, and ≥0.75 as excellent reproducibility (14). The significance level was set at P < 0.05.

Ethics

The study was approved by the local ethics committee, and all participants signed the informed consent form.

Results

s-Cortisol after cosyntropin test in healthy subjects

To define a LC-MS/MS–based cutoff value for s-cortisol 30 and 60 minutes after administration of cosyntropin, we enrolled and tested 121 healthy subjects not using oral estrogens. The median s-cortisol values at time 0 and at 30 and 60 minutes were 278 nmol/L (range, 108 to 653 nmol/L), 538 nmol/L (range, 395 to 843), and 619 nmol/L (range, 448 to 990 nmol/L), respectively. There were no obvious outliers (15). The 2.5th centiles for s-cortisol were 412 nmol/L (90% CI, 397 to 428) and 485 nmol/L (90% CI, 448 to 503) at 30 and 60 minutes, respectively. The 2.5th centiles for the increase in s-cortisol were 101 nmol/L (90% CI, 31 to 125) and 163 nmol/L (90% CI, 142 to 208) from 0 to 30 minutes and from 0 to 60 minutes, respectively. The median total increase was 325 nmol/L (range, 142 to 581 nmol/L). All but one healthy control subjects had a further increase in s-cortisol from 30 to 60 minutes, with a median value of 76 nmol/L (range, 17–202 nmol/L). The results are summarized in Table 2.

Table 2.

s-Cortisol and s-17-OHP Response in Healthy Control Subjects

	s-Cortisol (nmol/L), Median (Range)			s-17OHP (nmol/L), Median (Range)			Baseline p-ACTH (pmol/L)
	Baseline	30 Min	60 Min	Baseline	30 Min	60 Min	Median (Range)	n
Healthy control subjects without estrogens	278 (108–653)	538 (395–843)	619 (448–990)	1.7 (0.2–5.3)	3.4 (1.3–11.5)	3.8 (1.4–12.5)	4.3 (1.5–15.1)	121
Men	297 (164–598)	557 (395–843)	626 (485–930)	2.2 (0.8–4.0)	3.6 (2.3–8.9)	3.9 (2.1–9.0)	5.3 (1.5–15.1)	54
Women	260 (108–653)	536 (404–800)	619 (448–990)	1.1 (0.2–5.3)	3.1 (1.3-11.5)	3.4 (1.4-12.5)	3.9 (1.7–11.1)	67
Healthy control subjects on estrogens	665 (140–1168)	969 (547– 1348)	1071 (655–1614)	0.5 (0.2–1.4)	2.1 (1.0–12.8)	2.6 (1.1–13.2)	3.6 (0.3–12.7)	17
Age 20–40 y	290 (113–644)	552 (395–780)	621 (448–809)	2.1 (0.4–5.3)	3.4 (1.3–8.2)	3.4 (1.4–8.9)	4.3 (1.5–11.1)	60
Age 41–60 y	278 (108–653)	535 (418–843)	612 (485–990)	1.3 (0.2–5.3)	3.5 (1.5–11.5)	3.9 (1.7–12.5)	4.1 (1.7–15.1)	52
Age 61–80 y	248 (148–349)	586 (431–653)	656 (510–771)	1.6 (0.3–2.9)	4.8 (3.1–10.3)	4.5 (3.3–9.1)	6.4 (2.8–8.1)	9
Time at start
8:00–9:00 am	336 (206–569)	569 (414–716)	647 (448–752)	1.6 (0.6–5.1)	3.2 (1.3–6.9)	3.4 (1.4–7.3)	5.1 (1.5–10.9)	35
9:00–10:00 am	290 (172–653)	537 (412–806)	618 (505–990)	2.4 (0.3–5.3)	3.9 (1.5–8.2)	4.2 (1.7–8.9)	4.1 (1.7–9.8)	27
10:00–11:00 am	272 (190–536)	530 (475–736)	604 (559–7861)	1.8 (0.3–3.9)	3.9 (2.9–11.5)	4.3 (2.9–12.5)	4.7 (2.8–10.3)	7
11:00 am–12:00 pm	207 (108–644)	511 (395–673)	602 (472–794)	2.4 (0.3–5.3)	3.9 (1.5–8.2)	4.3 (2.9–12.5)	4.7 (2.8–10.3)	30
After 12:00 pm	265 (164–435)	533 (445–843)	610 (502–930)	1.5 (0.3–3.1)	3.5 (1.7–8,9)	3.9 (2.1–9.0)	5.0 (1.8–15.1)	19

Abbreviations: ACTH, adrenocorticotropic hormone.

Impact of sex, age, time of test, and kidney function

We found no significant difference in cortisol response between men and women or between age groups (Table 2). The response in cortisol was also evaluated according to what time of day the test was performed. The start time of the test was divided into five time intervals (Table 2). There was a tendency for a lower cortisol response at a later start time, although the only significant difference in cortisol response was found between 8:00 am and 9:00 am and between 11:00 am and 12:00 pm (P = 0.02 after 30 minutes; P = 0.014 after 60 minutes). There were no correlations between s-cortisol response and kidney function (measured as creatinine/estimated glomerular filtration rate), body mass index (BMI), and alcohol or smoking habits (not shown).

Seventeen healthy women on oral estrogens had a median baseline cortisol value of 665 nmol/L (range, 140 to 1168); 10 of these subjects had a basal cortisol >500 nmol/L. The median increase in cortisol during the test (from 0 to 60 minutes) was 414 nmol/L (range, 134 to 695), which was significantly higher than in healthy subjects without oral estrogens (P < 0.01). Two of the 17 women showed increases in cortisol of <101 nmol/L after 30 minutes or 163 nmol/L after 60 minutes.

Reproducibility of the test

The cosyntropin test was repeated in 20 healthy subjects. The mean time interval between the repeated tests was 256 (range, 27 to 385) days. Postcosyntropin cortisol levels at the two occasions revealed a strong positive correlation at 30 minutes (ρ = 0.90; P < 0.01) and at 60 minutes (ρ = 0.88; P < 0.01). The within-subject reproducibility in terms of ICC was 0.90 (95% CI, 0.97 to 0.99) and 0.97 (95% CI, 0.93 to 0.99) after 30 and 60 minutes, respectively.

Healthy control subjects evaluated according to the old cutoff for a positive test

Altogether, 65 of 121 (54%) healthy control subjects had s-cortisol <550 nmol/L, and 37 of 121 (31%) had s-cortisol ≤500 nmol/L 30 minutes after stimulation with cosyntropin. After 60 minutes, 21 of 121 (17%) subjects had s-cortisol <550 nmol/L, and 5 of 121(4%) had s-cortisol <500 nmol/L (Fig. 1). The five healthy subjects who had an s-cortisol <500 nmol/L after 60 minutes were further clinically and biochemically examined. Examination revealed no symptoms or signs indicative of AI. Repeated morning samples revealed p-adrenocorticotropic hormone in the normal range and s-cortisol >300 nmol/L. None of the subjects had autoantibodies against 21-hydroxylase. This practically ruled out PAI and SAI, and we found no indication for retesting any of the healthy subjects failing the old threshold with the cosyntropin test.

Figure 1.

s-Cortisol response to cosyntropin in healthy individuals.

Cortisol levels by immunoassays

To compare our results with the currently used immunoassays, 54 tests from healthy subjects were analyzed by the Immulite 2000 and Roche II methods. The 2.5th centiles were calculated in the same manner as for LC-MS/MS and are given in Table 3. The median cortisol values were higher at all time points for immunological assays than for LC-MS/MS (Table 3). The correlation coefficients between LC-MS/MS and Immulite 2000 were 0.95 at baseline, 0.85 at 30 minutes, and 0.87 at 60 minutes. The corresponding coefficients for Roche II were 0.98, 0.86, and 0.94, respectively. Correlation coefficients between Immulite 2000 and Roche II were 0.97, 0.83, and 0.86 at 0, 30, and 60 minutes, respectively. All correlations were significant (P < 0.01). Bland-Altman plots illustrate the variability between LC-MS/MS and the two immunoassays (Fig. 2) and reveal a somewhat larger variability for the Immulite assay than the Roche assay compared with LC-MS/MS.

Table 3.

Cortisol Measured With LC-MS/MS vs Immulite 2000 and Roche II in Healthy Control Subjects Without Oral Estrogens

	LC-MS/MS (n = 121)	Immulite (n = 49)	Roche (n = 49)
Median cortisol, nmol/L (range)
Baseline	278 (108–653)	287 (177–607)	304 (166–586)
30 min	538 (395–843)	582 (411–750)	590 (433–899)
60 min	619 (448–990)	646 (502–853)	674 (545–1024)
2.5th percentile for cortisol, nmol/L (90% CI)
30 min	412 (397–428)	423 (411–469)	440 (433–467)
60 min	485 (448–503)	504 (502–542)	548 (545–561)

Figure 2.

Bland-Altmannplot of immunological vs LC-MS assays of cortisol. The figure shows that both immunological methods on average measure higher values for cortisol than LC-MS/MS. Roche II seems slightly more accurate than Immulite 2000, based on narrower standard deviation.

17-OHP after cosyntropin test in healthy control subjects

The median 17-OHP values in the healthy control subjects not taking estrogens at 0, 30, and 60 minutes are shown in Table 2. No outliers were detected (15). The 97.5th percentile was 8.9 nmol/L (90% CI, 7.2 to 11.5) after 30 minutes and 9.0 nmol/L (90% CI, 8.3 to 10.2) after 60 minutes. Receiver operating curves shows that a cutoff after 60 minutes of 9.0 nmol/L gives 100% sensitivity for 17-OHP and 97% specificity for NCCAH. The median total increase was 1.8 nmol/L (90% CI, −0.6 to 11.6), and the 97.5th percentile for the increase in s-17-OHP from 0 to 60 minutes was 6.75 nmol/L (90% CI, 5.2 to 11.6). The basal s-17-OHP value was significantly lower in women than in men (P < 0.01). The response was significantly higher in men after 30 minutes (P = 0.03) but not after 60 minutes (P = 0.12). There were also differences between age groups. Basal 17-OHP was significantly lower in the elderly subjects (P < 0.01), but the response to cosyntropin was significantly higher in the elderly subjects both at 30 (P = 0.02) and 60 minutes (P = 0.02). The time of day did not influence the test result for s-17-OHP (Table 2). s-Creatinine/estimated glomerular filtration rate, BMI, and smoking and alcohol habits were not significantly correlated with postcosyntropin s-17-OHP (not shown). Alcohol consumption was low, with a mean intake of two units per week, and only 5 of 121 control subjects were smokers. Healthy subjects on oral estrogens had significantly lower s-17-OHP levels in basal blood (P < 0.01) and at 30 minutes (P = 0.02) and 60 minutes (P < 0.01) after cosyntropin stimulation (Table 2).

Reproducibility of 17-OHP assays

Post-cosyntropin s-17-OHP levels for test 1 and 2 correlated strongly both at 30 (ρ = 0.92; P < 0.01) and 60 minutes (ρ = 0.87; P < 0.01). The within-subject reproducibility in terms of ICC was 0.91 (95% CI, 0.77 to 0.96) after 30 minutes and 0.93 (95% CI, 0.83 to 0.98) after 60 minutes.

Patients with suspected AI

Of the 60 patients tested for suspected AI, 26 were subsequently diagnosed with AI, of whom six had PAI and 20 had SAI. The causes of AI are shown in Table 4. Median plasma adrenocorticotropic hormone at the start of the test was 72.8 pmol/L (95% CI, 9.9 to 278.0) in the suspected PAI group and 3.8 pmol/L (95% CI, 1.1 to 14.6) in the suspected SAI group. Patients diagnosed with PAI had a median increase in cortisol of 16 nmol/L (95% CI, −16 to 47) after 30 minutes and 15 nmol/L (95% CI, −36 to 86) after 60 minutes. Among patients tested for but not diagnosed with PAI, the corresponding values were 232 nmol/L (95% CI, 75 to 381) after 30 minutes and 276 nmol/L (95% CI, 147 to 483) after 60 minutes. Only one of the healthy subjects failed the 101 nmol/L (30 minutes) and 163 nmol/L (60 minutes) thresholds. All patients tested for SAI and found to be healthy had an increase in cortisol above these threshold values, as did 13 of 20 patients diagnosed with SAI according to the regular cutoffs.

Table 4.

Overview of Test Results in Patients With Suspected AI

Indication for Cosyntropin Testing	Suspected AI, n	Diagnosed AI, n
Total	60	26
PAI	14	6
SAI	46	20
Pituitary adenoma	9	4
Insufficiency due to the use of glucocorticoids	14	9
Long-acting opiate treatment	2	1
Unilateral adrenalectomy for autonomous cortisol secretion	6	5
Reason unknown	15	1

Of 60 patients with suspected AI, 22 (35%) had a cortisol value between 412 and 550 nmol/L 30 minutes after stimulation with cosyntropin, and 13 (22%) had a cortisol level between 412 and 500 nmol/L at the same time point. After 60 minutes, 7 of 60 patients (13%) had cortisol values between 485 and 550 nmol/L, and one patient had cortisol values 485 and 500 nmol/L after 60 minutes. These patients could be misdiagnosed with AI using conventional cutoff values.

Patients with suspected NCCAH

Thirty-four patients were tested for suspected NCCAH. The test was positive in three patients, of whom two had known NCCAH and one was diagnosed through our evaluation by positive cosyntropin test and genetic testing. The patients with verified NCCAH had significantly higher stimulated 17-OHP than healthy subjects. The cosyntropin-stimulated 17-OHP levels after 60 minutes were 12.1, 13.9, and 33.6 nmol/L, respectively, for the three patients with NCCAH.

Discussion

We demonstrate that a LC-MS/MS–based cortisol assay for the cosyntropin test is highly reproducible and lowers the cutoff values for a positive test to 412 and 485 nmol/L at 30 and 60 minutes, respectively. Recalculating cutoff levels for currently used immunoassay also lowers threshold values at 30 minutes but not at 60 minutes. The correlation between LC-MS/MS and the two most widely used immunoassays were best at baseline and decreased slightly after stimulation with cosyntropin. To apply correct threshold values is crucial to avoid overdiagnosing AI because the consequence can be life-long unnecessary replacement therapy.

Because we did not find significant differences in cortisol response between men and women or between different age groups, it is not necessary to provide age- and sex-specific reference ranges in adults, in line with previous findings (16–19). Moreover, we did not find that the time of day had a big impact, although levels were slightly higher when the test was performed early in the morning compared with noon.

A substantial proportion of healthy control subjects did not reach the cortisol cutoff level defining a normal test according to current guidelines. This highlights the importance of applying updated normal ranges and clinical action levels with a LC-MS/MS–based assay. The within-subject and between-subject correlations obtained by retesting 20 healthy individuals show that the reproducibility of the test is excellent (14). Notwithstanding, we recommend repeating a test with a suboptimal but near-normal result before introducing lifelong replacement with glucocorticoids.

There were no significant correlations between the cortisol response after cosyntropin and BMI, creatinine, and alcohol or smoking habits. This is also in agreement with previous findings (16). However, our population had overall normal kidney function and BMI, low alcohol consumption, and very few smokers.

In pregnant women and women on oral estrogens, interpretation of the test is not straightforward due to increased levels of cortisol-binding globulin (16, 20). An alternative approach to cutoff in these cases is to determine the increase in cortisol during the test. An increase of 200 to 250 nmol/L has typically been considered as a normal response to cosyntropin stimulation (21). Our data reveal that an even smaller increase in s-cortisol (101 nmol/L after 30 minutes and 163 nmol/L after 60 minutes) is indicative of an adequate cortisol response. However, we also found that cortisol increment is not sufficient for interpretation of the test because several patients with SAI had an increase in cortisol >200 nmol/L after both 30 and 60 minutes, and some healthy subjects failed to show such an increase. Healthy subjects on oral estrogens had a significantly higher increase in cortisol during the test compared with subjects without oral estrogens. However, this finding is uncertain because there were only 17 control subjects using estrogens.

Even though mass spectrometry is considered the gold standard for steroid analyses, immunological methods for cortisol measurements are still in widespread use (16, 17, 22). Current assays are much more specific than those used before. In agreement with the use of more specific antibodies, the Roche II and Immulite 2000 methods have lower cutoff values for the cosyntropin test after 30 minutes. However, based on our data, it is not possible to recommend specific cutoff levels for these methods because the number of tests analyzed by immunological methods was too low for accurate percentile calculations. The correlations between the different methods are excellent for baseline cortisol and are slightly lower, but still adequate, after 30 and 60 minutes. This fits well with the variance reported between different immunoassays (17), possibly due to the use different antibodies with different cross-reactivities. Cross-reactivities between steroids could also explain why immunological assays show higher average values for s-cortisol at all time points, especially at 30 and 60 minutes (Fig. 2). By using LC-MS/MS–based assays, interference between steroids is avoided, and these assays can be traced to international reference materials. Taken together, our results show that immunological tests are accurate and similar to LC-MS/MS but have median higher cortisol values possibly due to cortisol intermediates giving some interference.

Using the commonly recommended (2, 3) cutoff of 500 nmol/L for cortisol after cosyntropin stimulation after 30 minutes, we found 13% false positives in a cohort with suspected AI. Thus, it is important to use the redefined cutoffs. Otherwise, these patients are at risk for receiving lifelong glucocorticoid replacement therapy with possible side effects despite having a well-functioning corticotropic axis.

Perhaps the most striking result is that 17-OHP values were much lower in healthy subjects when assayed by LC-MS/MS compared with immunoassays. We show that a stimulated 17-OHP >9.0 nmol/L after 60 minutes, which is far from the current poststimulation cutoff of 30 to 45 nmol/L, is suggestive of NCCAH. Even our patients with verified congenital adrenal hyperplasia (n = 3) did not reach that level. However, because we were only able to test a few patients, further studies are needed to see if the cutoff should be adjusted. Our hospital receives referrals of patients with hirsutism, oligomenorrhea, and slightly elevated 17-OHP in basal blood samples. These patients could have low graded defects in adrenal enzymes, which will not be detected if current cutoffs are used. Furthermore, women showed lower basal measurements for 17-OHP than men, but there was no sex difference 60 minutes after cosyntropin. In elderly subjects, the 17-OHP response to cosyntropin was increased compared with younger subjects, but the significance of this finding is uncertain because the number of healthy subjects from the oldest age group was low.

In this setting of the cosyntropin test, the importance of specificity is even more important for 17-OHP than it is for cortisol. Decreased activity in one of the steroidogenic enzymes leads to a surge in upstream intermediates, increasing the risk for cross-reactivity in immunoassay methods. The results suggest that the test can be performed any time between morning and noon. Because both basal and stimulated 17-OHP values were significantly lower in subjects on oral estrogens, we recommend discontinuing estrogens before testing for NCCAH to avoid false-negative results.

In summary, our study identifies cutoff levels for cortisol and 17-OHP measured by LC-MS/MS that could significantly improve the diagnostic accuracy of adrenal failure. We find that the cutoff for cortisol after stimulation with cosyntropin should be moderated to 412 nmol/L after 30 minutes and 485 nmol/L after 60 minutes using LC-MS/MS. Failing to use proper cutoff levels could lead to misdiagnosis of AI in a substantial proportion of individuals. When the cosyntropin test is used to diagnose NCCAH, a 17-OHP level >9.0 nmol/L after 60 minutes should be considered the cutoff for a positive test.

Abbreviations:

Abbreviations:
 
• 17-OHP

  17-hydroxyprogesterone

 
• AI

  adrenal insufficiency

 
• BMI

  body mass index

 
• CI

  confidence interval

 
• CV

  coefficient of variation

 
• ICC

  intraclass correlation coefficient

 
• LC-MS/MS

  liquid chromatography tandem mass spectrometry

 
• NCCAH

  nonclassical congenital adrenal hyperplasia

 
• PAI

  primary adrenal insufficiency

 
• s-17-OHP

  serum 17-hydroxyprogesterone

 
• s-cortisol

  serum cortisol

 
• SAI

  secondary adrenal insufficiency

Acknowledgments

We thank Nina Jensen, Mona Eliassen, Gro M. Olderøy, Elisabeth T. Halvorsen, and Inger Svendsen for handling samples and registration forms.

Financial Support: The study was funded by grants from the University of Bergen, Western Regional Health Authorities, and Department of Medicine and Hormone Laboratory, Haukeland University Hospital.

Clinical Trial Information: ClinicalTrials.gov no. NCT0218660 (registered 14 June 2016).

Disclosure Summary: The authors have nothing to disclose.