Authors: Ken Kanazawa, Mai Hijikata, Koichiro Kuwabara
Categories: Original Research, adrenal insufficiency, actual clinical practice, hypothalamic-pituitary-adrenal axis, endocrine stress test
Source: AACE Endocrinology and Diabetes
Authors: Ken Kanazawa, Mai Hijikata, Koichiro Kuwabara
Adrenal insufficiency (AI) is characterized by glucocorticoid deficiency, with symptoms such as fatigue and hypoglycemia. Since AI presents with various severities in actual clinical settings, its evaluation using endocrine stress tests also requires a graded and comprehensive interpretation.
We aimed to classify AI severity in patients who underwent endocrine stress tests and analyze their clinical characteristics using real-world medical data.
This cross-sectional study was conducted at Tokyo Rosai Hospital.
Inpatients who underwent endocrine stress tests, including the insulin tolerance test, corticotropin-releasing hormone stimulation test, and standard-dose and low-dose corticotropin stimulation tests.
Patients were classified into 3 groups based on hormone overt AI, suspected early-stage AI, and normal adrenal function. Clinical characteristics, laboratory findings, and treatment interventions were analyzed across groups.
The 104 patients (mean 46.4 ± 13.5 years, 18.9% male, body mass 24.3 ± 6.7) exhibited low cortisol levels and fatigue. Among 90 patients completing severity-classifying tests, 19 (21.1%) had overt AI, 38 (42.2%) had suspected early-stage AI, and 33 (36.7%) had normal adrenal function. Corticotropin-releasing hormone stimulation test results indicated varying pituitary-adrenal axis function based on AI severity.
In routine care, a pragmatic classification combining stress tests can help stratify suspected cases. These findings underscore the importance of flexible and comprehensive interpretation of dynamic endocrine stress tests to detect diverse and subtle dysfunctions of the hypothalamic-pituitary-adrenal axis.
Highlights •This real-world cross-sectional study classified adrenal insufficiency (AI) severity—including early-stage AI—using combined endocrine stress testing•Early-stage AI was identified in over one-third of patients•The combined use of low-dose adrenocorticotropin stimulation test and corticotropin-releasing hormone stimulation test were shown to be valuable in suspecting early-stage AI•A flexible and comprehensive interpretation of dynamic endocrine test allow for the differentiation of diverse subtle functional disorders of the hypothalamic-pituitary-adrenal axis Clinical RelevanceThis study supports early detection of hypothalamic-pituitary-adrenal dysfunction using dynamic endocrine tests in clinical practice.
Adrenal insufficiency (AI) is an endocrine disorder characterized by absolute or relative glucocorticoid deficiency.1, 2, 3, 4, 5 This condition manifests with symptoms such as fatigue, reduced energy levels, loss of appetite, hypotension, and weight loss. Additionally, approximately 83% of affected patients experience severe hypoglycemia.^6^ However, because patients often remain asymptomatic or present with only nonspecific symptoms until the hypothalamic-pituitary-adrenal (HPA) axis fails to respond adequately to stress, AI is typically diagnosed at an advanced stage.^7^
A notable example of early-stage AI is latent, a hidden but potentially serious condition with an impaired response to corticotropin.^8^ However, patients with latent adrenal insufficiency generally lead normal lives with minimal or no clinical manifestations. Consequently, there are many cases of misdiagnosis, increasing the risk of life-threatening adrenal crises in response to stressful situations such as infections, overexertion, or surgery.^9^
The clinical spectrum of AI ranges from mild dysfunction to adrenal crisis,10, 11, 12 making timely and accurate diagnosis challenging. Selecting and evaluating appropriate endocrine stress tests is crucial, particularly in patients with early-stage AI.
Previous studies on early-stage AI have assessed the risk of developing AI in healthy individuals with positive organ-specific autoimmune diseases and adrenal autoantibodies by performing endocrine stress tests.13, 14, 15 However, this evaluation involved a single endocrinological stress test conducted on a target group within a research protocol and did not utilize data from endocrinological stress tests performed in actual clinical practice to reflect real-world patterns. In actual clinical practice, to diagnose AI across a broad spectrum of disease severities, including early-stage AI, it is necessary to discuss the graded evaluation and comprehensive interpretation of the results of multiple endocrine stress tests.
This study aimed to evaluate the dynamics of the HPA axis in patients who had undergone endocrine stress tests, classify AI severity, including early-stage AI, and analyze their clinical characteristics. The results of this study may provide important insights into AI management and ultimately improve the quality of life in patients.
This single-center cross-sectional study was conducted at the Department of Diabetes, Metabolism and Endocrinology, Japan Labor Health and Safety Organization, Tokyo Rosai Hospital. We reviewed the electronic medical records of patients who underwent a comprehensive endocrine stress test to investigate AI between January 2015 and November 2024. This study adhered to the World Medical Association Declaration of Helsinki and was approved by the Institutional Review Board of Tokyo Rosai Hospital (REC No. 06-35). This non-interventional database study did not require written informed consent, as per Japan's Ministry of Health, Labor, and Welfare guidelines. Patients were informed of the study via a website with the option to opt out.
We evaluated patients aged >18 years who underwent specific endocrine stress tests for suspected AI and were hospitalized in the endocrinology unit of our institution. Entry was not limited to low morning cortisol; indications included clinical suspicion (eg, fatigue, unexplained hypoglycemia) and/or screening laboratory abnormalities, as detailed in Table 1. During hospitalization, patients underwent early morning fasting blood sampling (06:00–08:00), 24-hour urine collection, insulin stimulation testing (ITT), standard-dose corticotropin stimulation test (SDST), low-dose corticotropin stimulation test (LDST), and corticotropin-releasing hormone testing (CRHST).Table 1Patient Characteristics, Indications for Endocrine Tests, Primary Clinical Complaints, and Comorbidities from the EMR DatabasePatient characteristicsPrimary clinical complaintsn (%)Comorbiditiesn (%)Number of patients retrieved from EMR database104Total74 (71.2)Total51 (49.0)Age, y45.4 ± 13.5Fatigue36 (34.6)Diabetes25 (24.0)Male sex, n (%)17 (18.9)Symptoms associated with hypoglycemia29 (27.9)Hypertension19 (18.3)BMI, kg/m^2^24.3 ± 6.7Nausea6 (5.8)Thyroid disease12 (11.5)Headache8 (7.7)Graves disease5 (4.8)Indications for endocrine testn (%)Dizziness6 (5.8)Chronic thyroiditis7 (6.7)Low cortisol levels in the early morning79 (75.9)Weight loss5 (4.8)Heart failure8 (7.7)Unexplained hypoglycemia6 (5.8)Others11 (10.6)Atrial fibrillation4 (3.8)Abnormal TSH levels4 (3.8)Old myocardial infarction4 (3.8)Abnormal ACTH levels2 (1.9)Respiratory disease8 (7.7)Pituitary abnormalities in MRI2 (1.9)Sleep apnea syndrome2 (1.9)Unexplained weight loss2 (1.9)Asthma6 (5.8)Severe menstrual pain2 (1.9)History of tuberculosis0 (0)Unknown7 (6.7)Dementia3 (2.9)History of cerebral infarction1 (1.0)Abbreviations: ACTH = adrenocorticotropin; BMI = body mass index; EMR = electronic medical records; MRI = magnetic resonance imaging.This table presents data on patient characteristics, indications for endocrine tests, primary clinical complaints, and comorbidities retrieved from the EMR database. Continuous variables are expressed as mean ± standard deviation, while categorical variables are presented as numbers (%).
Patients with any of the following characteristics were current diagnosis of malignancy, acute infection, recent surgery or severe trauma, excessive alcohol consumption, severe liver disease, nephrotic syndrome, pregnancy, and severe malnutrition. Demographic data (sex and age), indications for endocrine tests, primary clinical complaints, comorbidities, and endocrine stress test results, including endocrine assessments, were collected and analyzed.
The primary aim of this study was to comprehensively evaluate the HPA axis based on the results of the stress tests and to classify the severity of newly developed AI. The secondary aim was to analyze the clinical characteristics and stress test results according to disease severity. As exploratory outcomes, we investigated the intervention rate and average dose of glucocorticoid replacement therapy (GCRT) after AI diagnosis, which was titrated by body weight or body surface area, following current guideline recommendations, as well as the presence or absence of adverse events, including adrenal crisis, during the follow-up period.
General blood tests, including serum cortisol and other endocrine parameters, were performed during hospitalization. A skilled nurse collected a fasting blood sample early in the morning between 00 and 00 am in a supine position. During hospitalization, the use of exogenous steroid preparations (oral, ointment, inhalation, etc.) was discontinued. We performed all endocrine stress tests using pre-attached catheters to minimize additional stress to the patients. For urinary free cortisol (UFC) measurement, a 24-hour urine sample was collected. The total urine volume was recorded and UFC levels adjusted based on the urine creatinine concentration.
Serum cortisol levels (μg/dL) were measured using a second-generation cortisol assay and an automated electrochemiluminescence immunoassay (Cobas e801® unit, Roche Diagnostics, Mannheim, Germany) with a <3% intra-assay coefficient of variation, a <5% inter-assay coefficient of variation (reference, 6.02-18.2 μg/dL [6-8 AM]), and a low detection limit of 0.036 μg/dL.^16^ Instead of radioactive isotope labels, fluorescent labels, chemiluminescent labels, and other labeling methods are used in automated platform analyzers. Second-generation immunoassays that utilize monoclonal antibodies demonstrate high accuracy across the entire measurement range and show strong consistency with liquid chromatography-tandem mass spectrometry, which is the gold standard for cortisol measurement.^16^^,^^17^ Cutoffs (18 μg/dL) reflect assay-aligned thresholds from prior literature and were applied uniformly because all measurements used the same platform during the study. Generalizability across platforms (e.g., LC-MS/MS) may vary. The assay methods, measuring equipment, test reagent manufacturers, and reference ranges for other endocrinological parameters are listed in Supplementary Table1.
We evaluated AI and classified its severity.^1^ We assessed HPA axis response using ITT which targeted a plasma glucose ≤45 mg/dL, with bedside glucose monitoring every 15 min. Predefined stop rules (persistent neuroglycopenia, arrhythmia, or staff concern) were applied under endocrinologist supervision. Patients who failed to achieve the hypoglycemic threshold were excluded from subsequent screening. Patients with peak cortisol <18 μg/dL on ITT were considered to have HPA axis dysfunction and classified as suspected overt AI, including hypothalamic causes.^18^^,^^19^ However, if the cortisol level was ≥18 μg/dL, the pituitary-adrenal axis was assessed.^2^ We comprehensively evaluated the pituitary-adrenal axis using CRHST, SDST, and LDST. Based on the results of previous studies, we determined the cutoff values as In CRHST, interpretation was based primarily on cortisol response. A peak adrenocorticotropin (ACTH) levels (≤30 pg/mL or ≤ twofold increase from baseline) after corticotropin-releasing hormone (CRH) stimulation were used as supportive information to evaluate pituitary reserve.^4^^,^^20^ A 60-minute cortisol level <18 μg/dL in SDST and a 30-minute cortisol level <18 μg/dL in LDST were suggested AI.^3^^,^^4^^,^^14^^,^21, 22, 23 To classify AI severity, we continually evaluated the pituitary-adrenal axis by analyzing the differences in responses to SDST and LDST. In other words, if both the 60-minute cortisol value in SDST and the 30-minute cortisol value in LDST were <18 μg/dL, the case was classified as overt AI. If only the 30-minute cortisol value in LDST was <18 μg/dL, the case was classified as suspected early-stage AI, consistent with partial adrenal reserve. This category included both early primary AI and early central AI, depending on accompanying ACTH and CRHST results. Specifically, cases with preserved SDST response but blunted cortisol response to CRH stimulation were interpreted as early-stage central AI, indicating partial impairment of ACTH secretion at the pituitary level. If both the 60- and 30-minute cortisol values were ≥18 μg/dL, it was classified as normal adrenal function.
Continuous variables with normal distribution are expressed as mean ± standard deviation and compared using Student’s t test. In contrast, continuous variables with a non-normal distribution are expressed as median (interquartile range) and two-group comparisons used Mann–Whitney U or t-tests as appropriate. Analysis of covariance was used to compare gradual changes in efficacy variables between the 2 groups. The significance level was set at P < .05. All tests were two-tailed. Statistical analyses were performed using JMP version 12 software (SAS Institute Inc., Cary, NC, USA).
One hundred and four patients who underwent specific endocrine stress tests during the study period were extracted from the electronic medical record database. As shown in Table 1, the mean age of the patients was 46.4 ± 13.5 years, 18.9% were male, and mean body mass index was 24.3 ± 6.7. The primary reasons for undergoing endocrine stress tests were low cortisol levels (75.9%) and unexplained hypoglycemia (5.8%). The most common symptoms were fatigue (34.6%), hypoglycemia symptoms (27.9%), and nausea (5.8%). Comorbid conditions were present in 49.0% of the patients, with diabetes (24.0%), hypertension (18.3%), and thyroid disease (11.5%) being the most prevalent.
As shown in Fig. 1, among the 90 patients who completed severity-classifying tests (ie, excluding 14 who did not reach ITT hypoglycemia), 10 (11.1%) had overt AI suspicion (defined as a peak cortisol level <18 μg/dL on ITT, indicating HPA axis dysfunction). The final AI severity classification was determined using SDST and LDST results, identifying 9 patients (10%) with overt AI suspicion, consistent with pituitary or primary AI, 38 (42.2%) with suspected early-stage AI, and 33 (36.7%) with normal adrenal function. Based on the above, a total of 19 patients (21.1%) were diagnosed with overt AI suspicion.Fig. 1AI severity classification based on the results of endocrine stress tests. Abbreviations: AI = adrenal insufficiency; CAI = central adrenal insufficiency; CRHST = corticotropin-releasing hormone testing; HPA = hypothalamic-pituitary-adrenal; ITT = insulin stimulation testing; LDST = low-dose corticotropin stimulation test; PAI = primary adrenal insufficiency; SDST = standard-dose corticotropin stimulation test.
Table 2 presents the blood sampling, endocrinological, 24-hour urinary cortisol, and pituitary magnetic resonance imaging data for each AI severity classification. A significant difference in the serum potassium level was observed between the normal adrenal function (4.1 ± 0.3 mEq/L) and overt AI (3.9 ± 0.4 mEq/L) groups (P = .0044). There were no significant differences in fasting blood glucose or HbA1c levels between the normal adrenal function (103.1 ± 35.4 mg/dL, 6.1 ± 2%) and overt AI groups (91.1 ± 17.3 mg/dL, 5.4 ± 0.5%), although a trend toward lower levels was observed. No significant differences were found in basal cortisol levels between the normal adrenal function (8.9 ± 3.9 μg/dL) and overt AI groups (6.3 ± 3.4 μg/dL). Additionally, the mean ACTH and DHEA-S levels did not differ significantly between the 3 groups.Table 2Comparison of Laboratory, Endocrinological, 24-Hour Urine Collection, and Pituitary MRI Findings by AI SeverityNormalEarly-stage AI suspOvert AIP-value∗P-value†Number of patients, n (%)33 (36.6)38 (42.2)19 (21.1)Laboratory data WBC, 10^2^ μL56.7 ± 18.263.3 ± 18.156.8 ± 19.2.14.98 Hemoglobin, g/dL13.2 ± 1.013.4 ± 1.412.6 ± 1.6.32.11 Eosinophil count, %3.1 ± 2.43.6 ± 2.53.3 ± 1.6.34.79 ALB, g/dL4.2 ± 0.34.2 ± 0.44.0 ± 0.6.84.08 Sodium, mEq/L140.4 ± 2.7140.9 ± 2.6140.6 ± 1.9.35.74 Potassium, mEq/L4.1 ± 0.34.0 ± 0.43.9 ± 0.4.0189.0044 eGFR, mL/min/1.73 m^2^76.0 ± 19.083.5 ± 19.783.1 ± 30.7.16.27 Glu, mg/dL103.1 ± 35.4113.3 ± 47.491.1 ± 17.3.27.28 HbA1c, %6.1 ± 2.05.9 ± 1.05.4 ± 0.5.56.11Endocrinological data (early morning) Cortisol, μg/dL8.9 ± 3.911.1 ± 10.36.3 ± 3.4.21.22 ACTH, pg/mL18.4 ± 14.219.8 ± 14.820.1 ± 19.1.69.71 DHEA-S, μg/dL101.5 ± 51.4105.8 ± 62.383.2 ± 54.3.17.28 PRA, ng/mL/h3.0 ± 5.82.8 ± 3.22.9 ± 4.4.89.98 PAC, pg/mL124.8 ± 85.3170.7 ± 175.3163.3 ± 140.9.18.35 TSH, μIU/mL2.6 ± 2.11.9 ± 1.34.0 ± 6.13.37.1424-h urine collection data Total urine volume, mL/d1694.5 ± 623.41712.6 ± 1174.81577.8 ± 674.6.93.66 UFC/Cr ratio, μg/g Cr39.5 ± 28.337.2 ± 17.029.9 ± 19.3.88.29Pituitary morphology in MRI Empty cella, n (%)3 (9.0)3 (7.9)0 (0) Rathke's follicle, n (%)3 (9.0)2 (5.2)1 (5.3) Pituitary adenoma, n (%)1 (3.0)2 (5.2)1 (5.3) Post pituitary surgery, n (%)0 (0)1 (2.6)2 (10.5)Abbreviations: ACTH = adrenocorticotropin; AI = Adrenal insufficiency; ALB = albumin; DHEA-S = dehydroepiandrosterone sulfate; eGFR = estimated glomerular filtration rate; Glu = glucose; MRI = magnetic resonance imaging; n.s. = not significant; PAC = aldosterone; PRA = plasma renin activity; UA = Urinary aldosterone; UFC = urine free cortisol; WBC = white blood cell.This table presents data from the study participants, categorized based on the results of the endocrine stress test for comparative analysis.Continuous variables are expressed as mean ± standard deviation, while categorical variables are presented as numbers (%).Urine samples are collected over 24 hour with the patient hospitalized and under a restricted sodium diet.∗Comparison of values between the normal adrenal function and suspected early-stage AI groups.†Comparison of values between the normal adrenal function and suspected AI groups.
Fig. 2 shows an evaluation of adrenal function using SDST and LDST. The mean peak cortisol value in SDST was 21.9 ± 3.8 μg/dL at 60 minutes, while the mean peak cortisol value in LDST was 17.3 ± 2.9 μg/dL at 30 minutes, showing differences in response after stimulation (Fig. 2 A). There was a correlation between the 60-minute cortisol levels in SDST and 30-minute cortisol levels in LDST, with no abnormal values observed (Fig. 2 B). The cortisol response in SDST/LDST varied by case, and patients were classified into the normal adrenal function, suspected early-stage AI, and suspected AI groups according to the protocol. The mean 60-minute cortisol levels in SDST and mean 30-minute cortisol levels in LDST were 22.4 ± 5.1/20.1 ± 1.8 μg/dL in the normal adrenal function group, 20.9 ± 2.1/15.6 ± 2.1 μg/dL in the suspected early-stage AI group, and 15.4 ± 2.4/14.4 ± 2.5 μg/dL in the suspected AI group.Fig. 2Adrenal function evaluation using SDST and LDST. This figure presents the results of adrenal function assessments using SDST and LDST. (A) This chart presents the mean cortisol levels before, 30 minutes after, and 60 minutes after corticotropin stimulation in SDST and LDST, respectively. (B) This figure plots the 60-minute cortisol value in SDST and 30-minute cortisol value in LDST for each case. If both values were < 18 μg/dL, the case was classified as suspected AI; if only the 30-minute value was < 18 μg/dL, it was classified as suspected early-stage AI; and if both values were ≥ 18 μg/dL, it was classified as normal adrenal function. Abbreviations: AI = adrenal insufficiency; LDST = low-dose corticotropin stimulation test; SDST = standard-dose corticotropin stimulation test. ∗Statistically significant difference in the mean 30-minute cortisol value in LDST, using Student's t test (P < .0001). †Statistically significant difference in the mean 60-minute cortisol value in LDST, using Student's t test (P < .0001). ‡Statistically significant difference in the mean 30-minute cortisol value in SDST, using paired Student's t test (P < .0001). §Statistically significant difference in the mean 30-minute cortisol value in LDST, using Student's t test (P < .001).
Fig. 3 illustrates the evaluation of the pituitary-adrenal axis using CRHST, SDST, and LDST. Among the 80 patients with peak cortisol levels in ITT (≥18 μg/dL), 55 were classified into the normal ACTH-secreting group and 25 into impaired ACTH-secreting group based on CRHST. In the normal ACTH-secreting group (Fig. 3 A), the mean 30- and 60-minute cortisol levels in CRHST varied depending on the SDST/LDST results. The mean peak cortisol level at 60 minutes in CRHST was significantly lower in the suspected early-stage AI and AI groups (16.6 ± 2.3/14.2 ± 1.3 μg/dL, P < .0001) than in the normal adrenal function group (19.8 ± 2.8 μg/dL).Fig. 3Evaluation of the pituitary-adrenal axis using CRHST, SDST, and LDST. This figure illustrates the evaluation of the pituitary-adrenal axis using CRHST, SDST, and LDST. CRHST results are used to classify patients into (A) the normal ACTH-secreting group and (B) the impaired ACTH-secreting group. The figures present the mean cortisol value before, 30 minutes after, and 60 minutes after CRH stimulation for the following normal adrenal function, suspected early-stage AI, and suspected AI. Abbreviations: AI = adrenal insufficiency; CRHST = corticotropin-releasing hormone testing; LDST = low-dose corticotropin stimulation test; PC = peak cortisol; SDST = standard-dose corticotropin stimulation test. ∗Statistically significant difference compared with the normal adrenal function group, using Student's t test (P < 0.05). †Statistically significant difference compared with the normal adrenal function group, using Student's t test (P < .0001). ‡Statistically significant difference compared with the suspected early-stage AI group, using Student's t test (P < 0.05).
Additionally, a significant difference was observed between the suspected early-stage AI and AI groups (P < .05). The proportion of patients with peak cortisol levels <18 μg/dL after stimulation also varied according to SDST/LDST results, with the early-stage AI (70%) and AI (100%) groups showing higher rates than the normal adrenal function group (19%). In the impaired ACTH-secreting group (Fig. 3 B), the mean 60-minute cortisol level was significantly lower in the suspected early-stage AI group (14.2 ± 1.7 μg/dL, P < .0001) than in the normal adrenal function group (17.6 ± 2.7 μg/dL). The proportion of patients with peak cortisol levels <18 μg/dL after stimulation also varied by SDST/LDST results, with the early-stage AI (70%) and AI (100%) groups showing higher rates than the normal adrenal function group (19%).
Of the 57 patients with the early-stage and overt AI, 42 (73.7%) received GCRT after diagnosis, with an average dose of 5.8 ± 5.1 mg/d. No cases of adrenal crisis were observed during the mean follow-up period of 40.5 ± 28.9 months. During follow-up, 2 of the 19 patients (10.5%) with overt AI and 9 of the 38 (23.6%) with suspected early-stage AI visited the emergency room because of gastrointestinal symptoms and fever of unknown origin. Additionally, 2 of the 33 healthy participants (6.1%) were diagnosed with new-onset AI.
Among 90 patients completing severity-classifying tests, we identified overt AI (n = 19, 21.1%), suspected early-stage AI (n = 38, 42.2%), and healthy status (n = 33, 36.7%). Importantly, this study highlights that the combined use of the LDST and where available, the CRHST may provide supportive physiological insight into pituitary ACTH reserve.
In this study, AI severity could be classified into 3 groups using dynamic testing, regardless of basal cortisol levels. Previous studies have reported the limited usefulness of morning serum cortisol levels in distinguishing between normal and impaired HPA axis function. In contrast, dynamic testing is not restricted in its ability to differentiate between these states.^18^ Abnormalities in ITT were observed in 11.1% of cases, suggesting dysfunction of the HPA axis, including hypothalamic abnormalities. ITT remains the most established test for evaluating HPA axis response to stress.^18^ The HPA axis is regulated by neuroendocrine neurons in the paraventricular nucleus of the hypothalamus.^24^^,^^25^ It is believed that hypoglycemia induced by insulin stimulates these neurons, which triggers the release of both CRH and arginine vasopressin, leading to strong stimulation of ACTH secretion. Therefore, it is widely accepted that a peak cortisol level <18 μg/dL indicates a high probability of AI, necessitating replacement therapy.^4^ However, ITT is contraindicated in the elderly and in those with seizure disorders or cardiovascular diseases that are resistant to treatment.^26^^,^^27^ Additionally, determining the appropriate insulin dosage for patients with diabetes or severe obesity is challenging.^19^ Furthermore, as mentioned above, ITT provides an extremely strong stimulus, making it inappropriate to use this test to evaluate early-stage abnormalities of the pituitary-adrenal axis, which is the focus of this study. Therefore, alternative dynamic tests are required.
In other words, when evaluating early-stage abnormalities of the pituitary-adrenal axis, it is necessary to consider the pharmacological characteristics of SDST and LDST. SDST is widely used to confirm AI. This test evaluates the maximum secretory capacity of the adrenal cortex after the administration of 250 mcg of synthetic ACTH (1-24) (Corticotropin; Tetracosactide).^28^^,^^29^ This test can detect reduced cortisol response to pharmacological doses due to adrenal cortex destruction or atrophy resulting from the loss of endogenous ACTH action.^3^ In contrast, LDST is not recommended for diagnosing primary AI because it does not provide better diagnostic accuracy than SDST,^3^ and slight errors in dilution and injection volume may occur.^30^^,^^31^ However, numerous studies have reported that LDST plays an important role in diagnosing early-stage or partial abnormalities of the pituitary-adrenal axis.^13^^,^^14^^,^^22^^,^^32^ Yamamoto et al summarized the pathology of latent AI, an early-stage form of Addison disease.^8^^,^^12^ When latent AI with varying adrenal reserve capacity could occur, LDST is recommended for patients with unexplained symptoms, stress-related health changes, nonspecific symptoms, or coexisting autoimmune endocrine diseases. Similarly, Laureti et al demonstrated the clinical usefulness of LDST for diagnosing early-stage AI in patients with 21-hydroxylase autoantibodies.^13^ Furthermore, Pura et al^14^ reported that a cortisol level of 18.1 μg/dL (500 nmol/L) at 30 minutes of LDST serves as a cutoff value for defining early-stage AI, regardless of age, body weight, or body surface area.^14^ In pituitary AI, the extent of damage to ACTH-secreting cells depends on the pituitary lesion and the duration of its progression, leading to a spectrum of conditions, including mild conditions.^33^ Previous studies have shown that LDST has comparable diagnostic accuracy to SDST.^3^^,^^31^^,^^32^ The underlying principle is that in cases of chronic endogenous ACTH deficiency, the acute reactivity of the adrenal zona fasciculata is reduced, making it unlikely to generate an appropriate cortisol response. Based on the above findings, LDST is particularly useful for evaluating AI in the early or partial stages of both primary and central AI.
A notable finding of this study is that cortisol levels after CRH stimulation exhibited different responses depending on LDST and SDST results. Although CRHST has low diagnostic sensitivity for central AI and is not widely used,^34^ it remains effective for evaluating the response of ACTH-secreting cells in the pituitary gland.^33^ In this study, the cortisol response after CRH stimulation was lower in the LDST-positive group than in the LDST-negative group, despite normal ACTH secretion. This finding supports the existence of early-stage primary AI and aligns with the study by Boscaro et al, which demonstrated the presence of early-stage Addison disease using CRHST.^35^ In our cohort, serum DHEA-S levels were slightly higher in the suspected early-stage AI group compared with the normal group despite similar ACTH levels. This may reflect preserved zona reticularis function or compensatory upregulation during the early phase of HPA-axis dysfunction. The non-significant decline in serum albumin levels observed in overt AI may suggest a chronic catabolic state and warrants further investigation in larger prospective cohorts.
In this study, 73.7% of patients diagnosed with AI received GCRT after diagnosis, predominantly with oral hydrocortisone (average dose 5.8 ± 5.1 mg/d). Low-dose GCRT may potentially suppress residual pituitary ACTH secretion in patients with partial AI, theoretically contributing to corticotroph atrophy and eventual progression to overt AI. Therefore, in suspected early-stage AI, careful titration and regular re-evaluation of adrenal function are warranted to avoid overtreatment while preventing adrenal crisis. Current guidelines provide recommendations on GCRT for AI, with dosages typically determined based on body weight, where a glucocorticoid dose of 10–20 mg/d is considered equivalent to the physiological daily cortisol secretion.^1^^,^^4^ However, recent reports have highlighted the harmful effects of excessive GCRT.36, 37, 38 There are no established guidelines for GCRT in early-stage AI. Although individualized dose adjustments are necessary, determining whether the prescribed dose is excessive or insufficient remains challenging. It is unclear whether the GCRT administered to the 9 patients (23.6%) in the suspected early-stage AI group, who later presented with unexplained gastrointestinal symptoms or fever, was appropriate.
This study comprehensively analyzed AI severity classification based on endocrine stress test data derived from electronic medical records and clarified the current state of diagnosis and clinical management. A systematic evaluation of real-world data from multiple endocrine stress tests demonstrated the existence of early-stage AI and highlighted the need to reconsider the role of LDST, which has traditionally been used as an auxiliary diagnostic tool for central adrenal insufficiency.^3^^,^^31^^,^^32^ Previous studies have reported the use of LDST as a means of identifying early-stage primary adrenal insufficiency.^8^^,^^13^^,^^14^^,^^22^ In this study, in addition to LDST, we incorporated CRHST to further elucidate the dynamics of the pituitary-adrenal axis, reinforcing the importance of LDST in diagnosing early-stage primary adrenal insufficiency. Additionally, this study raises important clinical questions regarding the appropriate application of GCRT for diagnosed early-stage AI, optimal dosage strategies, and the extent of symptom improvement.
This study had some imitations. First, since this was a single-center, retrospective observational study, the findings may not be generalizable to other populations or healthcare settings. Second, selection bias is inherent in the study design, because we included only patients who underwent endocrine stress testing. This inclusion criterion may have led to an overestimation of the prevalence of AI owing to the pre-test probability of finding AI in this selected population. Third, the lack of long-term follow-up precludes a comprehensive assessment of the progression of early-stage AI and its clinical significance over time, underscoring the need for further longitudinal studies. Additionally, the potential variability in cortisol assay methods may have influenced the diagnostic evaluation of AI. Although liquid chromatography-tandem mass spectrometry is considered the gold standard owing to its high specificity for cortisol molecules, its application remains technically challenging and is not widely used in clinical practice.^23^^,^^39^ Second-generation cortisol immunoassays, which are commonly employed in routine diagnostics, demonstrate good accuracy.^16^ Assay-specific cutoffs may limit external comparability; however, internal comparisons remain valid given a single platform across the cohort. However, serum cortisol measurements are subject to limitations due to various confounding factors, including specific medications, comorbidities, sex, and age, all of which can impact cortisol levels and their interpretation across different assay platforms.^16^^,^^40^^,^^41^ Furthermore, we could not fully account for confounding factors that influence the HPA axis, such as the regular use of complementary medicine, physical inactivity, sleep disturbances, and night-shift work. Finally, the impact of corticosteroid-binding globulin was not considered in our analysis. While we excluded conditions known to cause a marked reduction in corticosteroid-binding globulin levels, such as liver cirrhosis, severe illness, and nephrotic syndrome, corticosteroid-binding globulin concentrations may be influenced by chronic stress, which plays a role in HPA axis regulation.42, 43, 44 Additionally, the cutoff thresholds for cortisol responses in LDST and CRHST are less standardized, and their abnormal results may overestimate partial AI prevalence, raising the possibility of overdiagnosis. Given these limitations, future studies involving a broader population, standardized diagnostic criteria, and long-term follow-up are warranted to enhance the clinical applicability of our findings.
This study challenges the traditional diagnostic paradigm for AI and proposes an evidence-based framework for redefining early-stage disease detection in endocrine practice. In particular, the combined application of LDST and CRHST may serve as a practical tool to support the diagnosis of early-stage AI in clinical practice. This simple stepwise approach could be readily adopted in secondary-care endocrine units.
Finally, this study highlights the importance of flexible and comprehensive interpretation of dynamic endocrine tests to detect diverse and subtle dysfunctions of the HPA axis. Future studies are needed to establish standardized guidelines for early-stage AI and improve patient outcomes.
The datasets generated and/or analyzed in this study are not publicly available. These data are available from the corresponding author upon request.
None disclosed.