Normalized Early Postoperative Cortisol and ACTH Values Predict Nonremission After Surgery for Cushing Disease

Abstract

Context:

Perioperative increases in adrenocorticotropic hormone (ACTH) and cortisol mimic results of corticotropin-releasing hormone (CRH) stimulation testing. This phenomenon may help identify patients with residual adenoma after transsphenoidal surgery (TSS) for Cushing disease (CD).

Objective:

To predict nonremission after TSS for CD.

Design:

Retrospective case-control study of patients treated at a single center from December 2003 until July 2016. Early and medium-term remission were assessed at 10 days and 11 months.

Patients and Setting:

Two hundred and ninety-one consecutive TSS cases from 257 patients with biochemical evidence of CD seen at a clinical center.

Interventions:

Normalized early postoperative values (NEPVs) for cortisol and ACTH were calculated as immediate postoperative cortisol or ACTH levels minus preoperative post–CRH-stimulation test levels.

Main Outcome Measures:

Prediction of early nonremission was evaluated using logistic regression. Prediction of medium-term remission was assessed using Cox regression. Predictive ability was quantified by area under the receiver operating characteristic curve (AUROC).

Results:

NEPVs for cortisol and ACTH predicted early nonremission [adjusted odds ratio (OR): 1.1; 95% confidence interval (CI): 1.0, 1.1; P = 0.016 and adjusted OR: 1.0; 95% CI: 1.0, 1.0; P = 0.048, respectively]. AUROC for NEPV of cortisol was 0.78 (95% CI: 0.61, 0.95); for NEPV of ACTH, it was 0.80 (95% CI: 0.61, 0.98). NEPVs for cortisol and ACTH predicted medium-term nonremission [hazard ratio (HR): 1.1; 95% CI: 1.0, 1.1; P = 0.023 and HR: 1.0; 95% CI: 1.0, 1.0; P = 0.025, respectively].

Conclusions:

NEPVs for cortisol and ACTH predicted nonremission after TSS for CD.

We studied normalized early postoperative values (NEPVs) for ACTH and cortisol and found that they improved prediction of nonremission in patients undergoing surgery for Cushing disease.

Persistent hypercortisolism is associated with immunosuppression, diabetes, hypertension, osteoporosis, and higher mortality rates (1). Transsphenoidal surgery (TSS) is the preferred treatment of hypercortisolism because of adrenocorticotropic hormone (ACTH)-secreting pituitary adenomas, or Cushing disease (CD). TSS has resulted in remission in 41% to 98% of CD patients (2–6) and has increased long-term survival (7–9).

In CD, preoperative corticotropin-releasing hormone (CRH) stimulation results in an ACTH response from the adenoma (10, 11) and not from the suppressed normal gland. Complete surgical resection of the pituitary adenoma is associated with postoperative hypocortisolism and suppressed CRH response as a result of long-standing corticotroph suppression in the normal gland (12). As a result, low postoperative cortisol and ACTH levels have been used as a marker of remission from CD, although considerable variability exists in published cutoff levels as well as in sampling times for establishing early remission (13–15). This variability is partly due to differences in tumor size, cavernous sinus invasion, the presence of an ectopic tumor, perioperative glucocorticoid use, corticosteroid-binding protein (transcortin) level, and the extent of tumor resection (16, 17). Postoperative basal cortisol levels have been used to predict long-term nonremission (2). However, newer methods are needed to decrease variability in early postoperative cortisol and ACTH levels and increase accuracy in the prediction of early nonremission.

When detected early, nonremission can lead to early reoperation. Patients benefit from early reoperation for nonremission following TSS for CD (18, 19). Thus, predictors of early nonremission can help identify patients who will benefit from early reoperation. We observed that within individuals, same-day postoperative (POD-0) laboratory cortisol and ACTH levels correlated with subsequent remission from CD (unpublished findings). We hypothesized that the stress from TSS presents an opportunity to analyze POD-0 results similar to a CRH stimulation (CRH-stim) test (12, 20) to predict remission from CD. Here, we present a strategy to minimize variability and improve prediction of nonremission using biochemical data available on POD-0 following TSS.

Materials and Methods

Patient data

This was a retrospective case-control study with patients in nonremission as cases and patients in remission as controls. Our cohort consisted of 257 patients who underwent 291 consecutive sublabial TSSs for ACTH-secreting pituitary adenomas by the same team of surgeons (E.H.O., R.R.L., and P.C.) at the Surgical Neurology Branch of the National Institutes of Neurological Disorders and Stroke at the National Institutes of Health (NIH) from December 2003 until July 2016 under research protocol ID 03-N-0164. Thirty-two patients underwent reoperations because of persistent CD, and two patients underwent a second reoperation. Written informed consent was obtained from each patient for research study participation. This study was approved by the National Institutes of Neurological Disorders and Stroke institutional review board.

Diagnosis of CD

Hypercortisolism was diagnosed on the basis of elevations in at least two of the following test values: late-night salivary cortisol (chemiluminescent enzyme immunoassay; Siemens Immulite 1000 analyzer, NIH Department of Laboratory Medicine, Bethesda, MD; normal range, <100 ng/dL), 24-hour urine free cortisol (high-performance liquid chromatography/tandem mass spectrometry, 2001 to 2005, Mayo Medical Laboratories, Rochester, MN; normal range: 8 to 77 μg per 24 hours; 2005 to present, NIH Department of Laboratory Medicine, Bethesda, MD; normal range: 1.4 to 20 μg per 24 hours for 3- to 8-year-olds, 2.6 to 37 μg per 24 hours for 9- to 12-year-olds, 4.0 to 56 μg per 24 hours for 13- to 17-year-olds, and 3.5 to 45.0 μg per 24 hours for ≥18-year-olds) or low-dose dexamethasone suppression testing (DST; 1 mg overnight or 2 mg over 48 hours; normal serum cortisol range: <1.8 μg/L) (21, 22).

Supplemental diagnostic testing included midnight serum cortisol (normal range: <7.5 μg/dL) and ACTH (Nichols Advantage Immunochemiluminometric Assay or chemiluminescence immunoassay; Siemens Immulite 2500 analyzer, NIH Department of Laboratory Medicine, Bethesda, MD; 2000 to 2005, normal range: 9 to 52 pg/mL; 2005 to 2015, normal range: 0.0 to 46.0 pg/mL; 2015 to 2016, normal range: 5 to 46 pg/mL) (23, 24) and 24-hour urinary 17-hydroxycorticosteroids excretion (1980 to 2007, NIH Department of Laboratory Medicine, Bethesda, MD; normal range: 3 to 10 mg per 24 hours for males and 2 to 6 mg per 24 hours for females; 2007 to 2016, Quest Diagnostics Nichols Institute, Chantilly, VA; normal range: 3 to 10 mg per 24 hours for males and 2 to 6 mg per 24 hours for females) (25). Pituitary source of ACTH was determined by early morning ACTH levels, 8 mg overnight DST, and ovine CRH-stim testing (26–28). Average post-CRH cortisol levels were determined by averaging serum cortisol levels at 30 and 45 minutes after CRH-stim; for ACTH, levels were determined by averaging serum levels at 15 and 30 minutes after CRH-stim. High-resolution (1- to 1.5-mm slice thickness) magnetic resonance imaging of the pituitary was performed to identify adenomas using spoiled gradient-recalled acquisition sequences, and 64-slice abdominal computed tomography scans were performed to assess for adrenal hyperplasia. Confirmatory inferior petrosal sinus sampling (29) was performed in all patients, except when DST and ovine CRH tests were concordant and magnetic resonance imaging showed an adenoma greater than 6 mm.

Postsurgical outcomes

Exogenous glucocorticoids were withheld during or after TSS until symptomatic hypocortisolemia was diagnosed. Postoperatively, serum cortisol and ACTH levels were routinely obtained at 6-hour intervals from POD-0 through day 3 (solid-phase competitive chemiluminescent enzyme immunoassay; Siemens Immulite 2500 Analyzer; NIH Department of Laboratory Medicine, Bethesda, MD; normal range: 5 to 25 μg/dL). On POD-0, first postoperative serum cortisol and ACTH levels were obtained at noon or 6 pm depending on the time of completion of surgery. After POD-3, daily morning cortisol levels were obtained until POD-10 or until patient discharge. Early remission was defined using well-established biomarkers, including nadir serum cortisol level of <5 μg/dL or <2 μg/dL within 10 days of surgery, before administration of exogenous glucocorticoids. Early reoperations were considered in patients with lack of resolution of hypercortisolism and lack of histopathologic evidence of ACTH-staining adenoma (18). Medium-term nonremission was confirmed at an average of 11 months of follow-up by clinical stigmata of Cushing syndrome, biochemical testing (24-hour urine free cortisol or cortisol circadian rhythm), need for medications targeting Cushing syndrome (e.g., ketoconazole), and/or requirement for further surgical management (20).

Statistical analysis

Patient baseline clinical characteristics are presented as percentages or as means with standard deviations. Baseline characteristics were compared using two-sample tests of proportions or two-sample t tests with unequal variance where appropriate, using Welch approximation for degrees of freedom (30). Agreement between early nonremission and medium-term nonremission was determined using unweighted κ statistics (31, 32). Predictors of early nonremission were identified by univariable and multivariable logistic regression with early nonremission as the dependent variable. Predictive models were evaluated for goodness-of-fit using AUROCs. Score cutoffs were chosen by sensitivity and specificity analyses after AUROC to maximize sensitivity and were confirmed using 2 × 2 tables. Standard errors were calculated by the DeLong method (33). Predictors of medium-term nonremission were identified by Cox regression, using the Breslow method for ties (34). Survival between groups was visualized using Kaplan-Meier curves. P values <0.05 two-tailed were considered statistically significant. GraphPad Prism 6.0 software was used for statistical figures (GraphPad Software, La Jolla, CA). Statistical analyses were performed using STATA 14/IC software package (StataCorp LP, College Station, TX).

Results

Two hundred and fifty-seven patients underwent 291 TSS procedures during this study. Two hundred and fifty-one procedures (86.3%) resulted in early postoperative hypocortisolemia with cortisol level <5 μg/dL, and 224 procedures (77%) resulted in early hypocortisolemia with cortisol level <2 μg/dL. Based on a cutoff of <5 μg/dL, patients who achieved early remission were significantly younger (26.4 vs 35.7 years; P = 0.002) but did not differ significantly in body mass index (BMI), sex, race, or percentage with pathology-confirmed adenomas (Supplemental Table 1). Patients in remission had significantly lower average daily serum ACTH and cortisol levels on all postoperative days tested, including as early as POD-0 (P < 0.05 each day; Supplemental Fig.; Supplemental Tables 2 and 3). In agreement with a previous report (14), both cutoffs (cortisol nadir <5 µg/dL or <2 µg/dL) appeared to distinguish equally between patients in remission and nonremission; therefore, we used the cutoff of <5 μg/dL in subsequent analyses. Of the 32 patients who underwent repeated TSS, 23 had early reoperations within 28 days for persistent disease. Of these, 21 patients (91.3%) achieved early remission. Therefore, when including early reoperations, 234 of 257 patients (91%) achieved remission on the basis of a cutoff of <5 μg/dL; this finding confirmed the utility of early reoperation as a successful strategy when persistent CD is diagnosed in the early postoperative period.

Next, we identified predictors of nonremission using logistic regression, with early nonremission as the dependent variable. In agreement with other studies (35), POD-1 cortisol levels predicted early nonremission. In addition, POD-0 cortisol levels at noon and at 6 pm significantly predicted early nonremission (P < 0.05). After adjustments for age, sex, and BMI, POD-0 cortisol levels at noon were no longer a significant predictor of early nonremission, most likely because of the limited numbers of patients with morning surgeries and complete data, whereas POD-0 at 6 pm, POD-1 at midnight, and POD-1 at 6 am remained significant predictors of early nonremission. Serum ACTH levels at POD-0 at 6 pm, POD-1 at midnight, and POD-1 at 6 am were significant predictors of early nonremission after multivariable analysis (Table 1).

Table 1.

Predictors of Early Nonremission

Variables	OR	95% CI	Standard Error	P Value	Adjusted OR	Adjusted P Value
Demographics
Age	1.0	1.0, 1.0	0.01	0.002	—	—
BMI	1.0	1.0, 1.1	0.02	0.097	—	—
Sex	0.7	0.4, 1.5	0.27	0.403	—	—
Path confirmed	0.5	0.2, 1.1	0.21	0.102	—	—
White	0.8	0.4, 1.7	0.30	0.570	—	—
Black	1.5	0.6, 3.7	0.70	0.364	—	—
Asian	—	—	—	—	—	—
Cortisol
Post-CRH % change	1.0	1.0, 1.0	0.00	0.604	1.0	0.614
Post-CRH avg	1.0	1.0, 1.0	0.02	0.685	1.0	0.721
Post-CRH max	1.0	1.0, 1.0	0.02	0.961	1.0	0.858
POD-0 at noon	1.0	1.0, 1.1	0.02	0.034	1.0	0.165
POD-0 at 6 pm	1.1	1.0, 1.1	0.02	<0.001	1.1	<0.001
POD-0 avg	1.1	1.0, 1.1	0.02	<0.001	1.1	<0.001
POD-1 at midnight	1.1	1.0, 1.1	0.02	<0.001	1.1	<0.001
POD-1 at 6 am	1.2	1.1, 1.2	0.03	<0.001	1.2	<0.001
ACTH
Post-CRH % change	1.0	1.0, 1.0	0.00	0.287	1.0	0.528
Post-CRH avg	1.0	1.0, 1.0	0.00	0.244	1.0	0.333
Post-CRH max	1.0	1.0, 1.0	0.00	0.213	1.0	0.309
POD-0 at noon	1.0	1.0, 1.0	0.00	0.054	1.0	0.082
POD-0 at 6 pm	1.0	1.0, 1.0	0.00	0.021	1.0	0.025
POD-0 avg	1.0	1.0, 1.0	0.00	0.005	1.0	0.007
POD-1 at midnight	1.0	1.0, 1.0	0.01	0.019	1.0	0.035
POD-1 at 6 am	1.0	1.0, 1.1	0.01	<0.001	1.0	0.001

Nonremission was defined as no postoperative cortisol level <5 μg/dL within 10 days of surgery. Results from univariable or multivariable logistic regression adjusting for age, sex, and BMI, with nonremission as the dependent variable.

Abbreviations: avg, average; max, maximum.

We further assessed the strength of association between postoperative cortisol or ACTH levels and nonremission using AUROC. Serum cortisol level on POD-1 at 6 am yielded the highest AUROC (0.96). Serum cortisol levels on POD-0 at noon and at 6 pm also predicted early nonremission, albeit with a lower AUROC (0.72 vs 0.96, P = 0.013; and 0.85 vs 0.96, P = 0.019; respectively; Table 2). Serum ACTH level at POD-0 or POD-1 did not predict early remission as well as POD-1 cortisol at 6 am (P < 0.05 each), most likely because of the high variability in serum ACTH levels compared with cortisol measurements (Table 2). Sensitivity and specificity analyses after AUROC identified a cutoff of 19.4 μg/dL for serum cortisol at POD-1 at 6 am. At this cutoff, sensitivity was 100% and specificity was 90% for predicting early nonremission.

Table 2.

Comparison of Postoperative Laboratory Test Results Predicting Early Nonremission

Variable	AUROC	95% CI	Standard Deviation	n	P Value
POD-0 at noon cortisol	0.72	0.55, 0.90	0.852	89	0.013
POD-0 at 6 pm cortisol	0.85	0.77, 0.93	0.456	118	0.019
POD-0 avg cortisol	0.85	0.76, 0.94	0.509	125	0.034
POD-1 at midnight cortisol	0.91	0.85, 0.96	0.297	115	0.122
POD-1 at 6 am cortisol	0.96	0.93, 0.99	0.188	142	—
POD-0 at noon ACTH	0.76	0.60, 0.92	0.718	80	0.018
POD-0 at 6 pm ACTH	0.73	0.60, 0.86	0.713	116	0.001
POD-0 avg ACTH	0.73	0.59, 0.86	0.771	123	0.002
POD-1 at midnight ACTH	0.74	0.63, 0.85	0.617	117	0.000
POD-1 at 6 am ACTH	0.82	0.72, 0.93	0.618	136	0.018
NEPV cortisol	0.78	0.61, 0.95	0.768	78	0.047
NEPV ACTH	0.80	0.61, 0.98	0.784	71	0.093

Results of AUROC analyses with early nonremission as the reference variable. Nonremission was defined as no postoperative cortisol level <5 μg/dL within 10 days of surgery. P values from two-sample t tests after testing for variance equality, with POD-1 at 6 am cortisol chosen as the control. NEPVs for cortisol and ACTH were calculated as the difference between serum levels at POD-0 at noon and average preoperative post–CRH-stim test values.

Because serum ACTH and cortisol levels are known to rise immediately after surgery during anesthesia reversal and extubation (16, 36), we investigated whether perioperative stress acts as an endogenous CRH-stim test and whether this can be used to improve prediction using POD-0 ACTH and cortisol levels. First, we assessed possible correlations between population-aggregated post–CRH-stim levels before surgery and POD-0 levels of serum cortisol and ACTH. In patients who did not achieve early hypocortisolism, postoperative levels of both ACTH and cortisol were not significantly different from those achieved during preoperative CRH-stim tests [Fig. 1(a) and 1(c)]. Conversely, in patients who achieved early hypocortisolism, postoperative ACTH and cortisol levels were significantly lower (P = 0.002 and P < 0.001, respectively) than post–CRH-stim levels before surgery [Fig. 1(b) and 1(d)]. These findings supported our hypothesis that surgical stress may act as an endogenous CRH-stim test and agree with an earlier report of decreased postoperative response to CRH-stim after successful surgery for CD (12).

Figure 1.

Comparison of CRH-stim testing before surgery vs postoperative serum (a, b) cortisol and (c, d) ACTH levels. Early remission was defined as at least one postoperative cortisol level <5 μg/dL within 10 days of surgery. P values are from comparison of post-CRH to POD-0 at noon laboratory values using two-sample t tests after testing for variance equality. Pre CRH and post CRH are preoperative. Horizontal bars represent mean ± SD. NS, nonsignificant; SD, standard deviation.

To take further advantage of this observation, we subtracted POD-0 levels of cortisol and ACTH at noon from average post–CRH-stim levels before surgery to generate normalized early postoperative values (NEPVs). NEPV data for cortisol were available for 78 early morning cases, and NEPV data for ACTH were available for 71 cases. Of these, NEPVs for cortisol and ACTH were significantly lower for patients who reached early remission than for those who did not [P = 0.001; Fig. 2(a) and 2(b)], and this was readily apparent in select cases with initial nonremission and successful early reoperation within 1 month [Fig. 2(c) and 2(d)]. Furthermore, after normalization using NEPV, early POD-0 noon cortisol and ACTH levels became significant predictors of early nonremission after adjusting for covariates (NEPV cortisol: OR, 1.1; 95% CI, 1.0, 1.1; P = 0.016 and NEPV ACTH: OR, 1; 95% CI, 1, 1; P = 0.048). NEPV for cortisol predicted early nonremission with AUROC, 0.78; 95% CI, 0.61, 0.95, and NEPV for ACTH predicted early nonremission with AUROC, 0.80; 95% CI, 0.61, 0.98 (Table 2), confirming improved predictive accuracy using NEPV. At a cutoff of −12, NEPV for cortisol predicted early nonremission with 100% sensitivity, 39% specificity, 100% negative predictive value, and 18% positive predictive value. A cutoff of −40 in NEPV for ACTH predicted early nonremission with 88% sensitivity, 41% specificity, 96% negative predictive value, and 16% positive predictive value. These results again likely reflect greater variability in ACTH measurements.

Figure 2.

Comparison of NEPVs for (a) cortisol and (b) ACTH between patients in early remission vs patients in nonremission. NEPV was defined as immediate postoperative cortisol or ACTH value (POD-0 at noon) minus average preoperative post–CRH-stim test values. Early remission was defined as at least one postoperative cortisol level <5 μg/dL within 10 days of surgery. Dashed lines at NEPV = 0. P values from two-sample t tests with unequal variance. Comparison of NEPVs for (c) cortisol and (d) ACTH between initial and repeated TSSs for patients undergoing early repeated surgery within 1 month. NEPV was defined as immediate postoperative cortisol or ACTH value (POD-0 at noon) minus average preoperative post−CRH-stim test values.

Finally, we assessed the utility of NEPV for predicting medium-term nonremission after an average of 11 months of follow-up. We observed 88.7% agreement between early and medium-term nonremission (unweighted κ, 0.63; P < 0.001), indicating substantial agreement (31). Accordingly, NEPV for both cortisol and ACTH levels predicted medium-term nonremission [hazard ratio (HR) 1.1; 95% CI: 1.0, 1.1; P = 0.023; and HR: 1.0; 95% CI: 1.0, 1.0; P = 0.025; respectively; Table 3]. Applying the cutoffs of −12 for NEPV for cortisol and −40 for NEPV for ACTH, we demonstrated good separation between patients above and below the cutoffs using Kaplan-Meier curves (Fig. 3). Thus, normalized early postoperative cortisol values may be used as a sensitive screening test to detect nonremission on the same day following TSS.

Table 3.

Predictors of Medium-Term Nonremission With Average Follow-Up Duration of 11 Months

Variables	HR	95% CI	Standard Error	P Value
Demographics
Age	1.0	1.0, 1.0	0.01	0.879
BMI	1.0	1.0, 1.1	0.02	0.248
Sex	1.2	0.5, 2.9	0.53	0.703
Path-confirmed	0.6	0.2, 1.6	0.29	0.280
White	0.9	0.4, 2.1	0.39	0.793
Black	1.2	0.4, 3.5	0.65	0.755
Asian	0.9	0.2, 4.0	0.68	0.855
Nadir	1.1	1.1, 1.2	0.03	<0.001
Nadir POD	1.1	1.0, 1.3	0.09	0.087
Cortisol
Post-CRH % change cortisol	1.0	1.0, 1.0	0.00	0.112
Post-CRH avg cortisol	1.0	0.9, 1.0	0.02	0.252
Post-CRH max cortisol	1.0	0.9, 1.0	0.02	0.141
POD-0 at noon cortisol	1.0	1.0, 1.1	0.01	0.070
POD-0 at 6 pm cortisol	1.0	1.0, 1.1	0.01	0.021
POD-0 avg cortisol	1.0	1.0, 1.1	0.02	0.008
POD-1 at midnight cortisol	1.0	1.0, 1.1	0.02	0.046
POD-1 at 6 am cortisol	1.1	1.0, 1.1	0.02	<0.001
NEPV cortisol	1.1	1.0, 1.1	0.03	0.023
ACTH
Post-CRH % change ACTH	1.0	1.0, 1.0	0.00	0.222
Post-CRH avg ACTH	1.0	1.0, 1.0	0.00	0.188
Post-CRH max ACTH	1.0	1.0, 1.0	0.00	0.128
POD-0 at noon ACTH	1.0	1.0, 1.0	0.00	0.077
POD-0 at 6 pm ACTH	1.0	1.0, 1.0	0.00	0.116
POD-0 avg ACTH	1.0	1.0, 1.0	0.00	0.026
POD-1 at midnight ACTH	1.0	1.0, 1.1	0.01	0.002
POD-1 at 6 am ACTH	1.1	1.0, 1.1	0.01	<0.001
NEPV ACTH	1.0	1.0, 1.0	0.00	0.025

Results from Cox regression with nonremission as the dependent variable.

Figure 3.

Kaplan-Meier curves comparing medium-term remission rates in patients above vs patients below NEPV cutoffs for (a) cortisol (NEPV −12) and for (b) ACTH (NEPV −40). Cutoffs from AUROC analyses used early nonremission as the classification variable. NEPV was defined as immediate postoperative cortisol or ACTH value (postoperative day 0 at noon) minus average preoperative post–CRH-stimulation test values.

Discussion

Treatment options after unsuccessful TSS for CD include repeated surgery, radiation, medical treatment, and bilateral adrenalectomy (37). Early reoperation has been associated with lower risk of altered anatomy due to scarring (5, 18, 19), but this practice is still debated because of the small subset of patients who may experience delayed remission (38). Accurate determination of early nonremission is thus of clinical importance because it may selectively identify patients who will benefit from early repeated surgeries.

The optimal time for determination of early remission has not been clearly established in the literature. Postoperative hypocortisolemia (<5 µg/dL), reduced plasma ACTH level (<10 pg/mL), or suppressed CRH response can accurately identify patients in remission from CD following TSS (4, 12, 14, 20). Hypocortisolemia and suppressed CRH response occur following successful elimination of CRH-responsive adenomatous corticotrophs (10, 11). The remaining nonadenomatous corticotrophs are in a state of long-standing suppression because of chronic hypercortisolism (39, 40), leading to profound postoperative deficits (41). Postoperative determination of hypocortisolism or suppressed CRH response is tested at least 24 hours after last corticosteroid administration. A CRH-stim test performed up to 10 days after surgery detects continued suppression of the hypothalamic-pituitary-adrenal (HPA) axis as a surrogate for remission from CD (12, 20). Likewise, serum hypocortisolemia or low ACTH levels within 12 hours to 5 days after surgery are reliable markers for remission from CD (4, 5, 12, 14, 18, 20, 42). We confirmed this observation in our cohort as well. POD-1 at 6 am serum cortisol level predicted remission accurately (AUROC, 0.96) (Table 3). Serum cortisol levels obtained earlier postoperatively (<12 hours) or intraoperatively lose their accuracy in predicting remission (13, 43–45). We found a similar decrease in the accuracy of POD-0 serum cortisol (AUROC, 0.72 to 0.85) and ACTH (AUROC, 0.73 to 0.76) levels for predicting early remission. We suspect that the reduction in accuracy is partly due to a transient spike in serum ACTH and cortisol levels following TSS for adenoma resection (16, 41, 44). In our cohort, regardless of the eventual remission status, initial spikes in serum cortisol and ACTH levels obfuscated the ability to predict eventual remission (Fig. 2).

Studies on the effect of perioperative stress on the HPA axis in non-CD patients offer clues that help interpret intraoperative and early postoperative hypercortisolemia (36, 46–48). In non-CD patients undergoing surgical procedures under general endotracheal anesthesia, maximum stimulation of the HPA axis with increased serum cortisol level occurs at the time of anesthesia reversal and endotracheal extubation (36, 47). Correspondingly, in CD patients, spikes in serum cortisol and ACTH levels are observed intraoperatively (45) and postoperatively (44). Perioperative stress can overcome acute HPA-axis suppression to cause hypercortisolemia (46). We believe that transient hypercortisolemia in patients who eventually achieve remission is due to the effect of surgical stress overcoming chronic HPA-axis suppression in nonadenomatous corticotrophs. This conjecture is supported by observations of delayed drops in serum ACTH and cortisol levels in instances of complete adenoma resection compared with hypophysectomy (41, 49), indicating continued ACTH secretion from nonadenomatous corticotrophs. At endotracheal extubation, acute activation of the HPA axis is most likely mediated by hypothalamic CRH release within the hypophyseal portal system (36). Endotracheal extubation typically occurs 20 to 30 minutes after removal of the adenoma (or exploration of sellar contents). This phenomenon therefore may be interpreted as an endogenous CRH-stim test following presumed adenoma resection. Corticotroph stimulation may also be mediated by interleukin-1 induced by surgical stress (50–52).

In this study, we took advantage of the observation that cortisol values tend to be consistent in the same individual relative to interindividual values (53). We normalized early postoperative cortisol levels to CRH-stim test results to generate NEPVs for POD-0 cortisol and ACTH levels (Fig. 2). In contrast to previous studies (13, 43–45), the analysis described here took advantage of using each patient as his or her own control, using the results of the cortisol and ACTH values immediately after surgery as the results of an endogenous CRH-stim test after the adenoma has (or has not) been completely removed. With complete adenoma removal, the endocrine responses should derive only from the nonadenomatous corticotrophs, whereas with residual adenoma the response is the combined effect of adenomatous and nonadenomatous corticotrophs. Early postoperative cortisol levels are relatively easy to obtain in the perioperative setting; therefore, their routine clinical use as a predictive tool potentially has broad applicability. The NEPV method minimizes variability arising from differential CRH responses across individuals and may be used in future studies to further evaluate the clinical utility of early postoperative cortisol and ACTH levels.

Despite predictive accuracy using NEPV, we reported the highest predictive accuracy using serum cortisol at 6 am on POD-1. This finding is partly due to greater data availability at the latter time (n = 142 vs 78). Indeed, the NEPV method was accurate in predicting remission on POD-0 at noon; however, its predictive ability was diminished by POD-0 at 6 pm and POD-1 at 6 am. The time window when NEPV shows the greatest benefit (i.e., <12 hours postoperatively) is precisely when population-aggregated serum cortisol levels have been historically unreliable in predicting remission (13, 43–45). Stricter control of immediate postoperative serum collection times may improve the predictive ability of NEPV. Currently, the first postoperative cortisol and ACTH levels are collected at approximately noon (clock time) following completion of the first TSS procedure of the day. For the second TSS procedures of the day, the first clock times for postoperative cortisol and ACTH collections are less uniform. A future study will examine whether NEPV similarly improves predictive accuracy at time of extubation. This may reduce variability (a limitation of the current study) by pegging sample collection times to surgery completion time. If validated, NEPV at extubation may obviate the need for multiday serial monitoring of cortisol and ACTH levels.

In this study, we reported an 86.3% rate of remission after TSS. Combining initial TSS and early repeated TSS, our early remission rate was 91% using a cutoff of <5 μg/dL and 83.5% using a cutoff of <2 μg/dL. Likewise, we confirmed an earlier report that cutoffs of 2 and 5 μg/dL for early remission appeared to similarly differentiate patients who achieved early remission from those who did not (14). We found statistically significant differences in postoperative averaged daily serum cortisol and ACTH levels between patients in remission vs those in nonremission (up to postoperative day 10) using either cutoff (Supplemental Tables 1 and 2). However, a cutoff of 5 μg/dL appeared to yield better separation between the remission and nonremission groups, with less overlap in their 95% CIs (Fig. 1) and was therefore used in the remainder of the study.

Our study had a number of limitations. We were unable to assess long-term remission or very early POD-0 laboratory values in a number of patients because of missing data, and this may have limited our ability to detect statistically significant differences and make long-term inferences. A more structured approach in collecting immediate postoperative serum cortisol and ACTH levels (i.e., at surgery completion) may improve the accuracy of the NEPV method for predicting nonremission. We were also unable to assess additional possible confounders because of unavailability of required data in this retrospective series. However, our study benefited from large numbers and from uniform procedures performed by the same surgeons and at the same center. Future studies will more fully explore the utility of very early postoperative laboratory values using a large prospective data set.

Summary and Conclusions

In this study, we evaluated the utility of early postoperative cortisol and ACTH levels for predicting nonremission after transsphenoidal adenomectomy for CD. POD-1 values at 6 am performed best at predicting early nonremission. POD-0 values at noon predicted early nonremission, albeit with a lower AUROC. Normalizing early cortisol and ACTH values to post-CRH values improved their prognostic value. Further prospective studies will explore the utility of normalized very early POD-0 cortisol and ACTH levels in identifying patients at risk for nonremission following TSS for CD.

Abbreviations:

ACTH

adrenocorticotropic hormone

AUROC

area under the receiver operating characteristic curve

BMI

body mass index

CD

Cushing disease

CI

confidence interval

CRH

corticotropin-releasing hormone

CRH-stim

corticotropin-releasing hormone stimulation

DST

dexamethasone suppression testing

HPA

hypothalamic-pituitary-adrenal

HR

hazard ratio

NEPV

normalized early postoperative value

NIH

National Institutes of Health

OR

odds ratio

POD

postoperative day

TSS

transsphenoidal surgery.