After obtaining the board approval of the local ethics committee, the medical files and imaging data of consecutive craniopharyngioma patients who underwent fully endoscopic endonasal approach (EEA) between December 2012 and March 2021 in our center were retrospectively reviewed. Inclusion criteria were listed as follows: (1) the diagnosis of craniopharyngioma was histologically confirmed; (2) medical records were complete, including demographic data, pre- and postoperative imaging data, preoperative assessment, tumor characteristics, intraoperative videos, surgical results and complications, postoperative management, follow-up, and endocrinological outcomes; and (3) endocrinological evaluation with a minimum 1-year follow-up was required. Patients undergoing reoperations were excluded because the possible initial manipulation of the HPA could affect the identification of the injury type of the HPA and the endocrinological outcomes. In addition, patients undergoing tumor recurrence or regrowth within 1 year of follow-up were excluded, because of possible effects on endocrinological outcomes. The manuscript conforms to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.

Tumor size estimated as the maximal tumor diameter, tumor location, tumor consistency, calcification, and hydrocephalus were determined by MR images and computed tomography (CT) scans before surgery. Regarding the extent of resection (EOR), gross total resection (GTR) was defined as no residual tumor evaluated by postoperative contrast-enhanced MR images acquired within 72 h after surgery. In contrast, cases in which ≥80% of the tumor was resected were deemed as subtotal resection (STR) (9).

Regarding the injury of the HPA, seven injury categories, including intact HPA (HPA-intact type; Figure 1A), exclusive pituitary stalk sacrifice (PSS) type (Figure 1B), hypothalamic injury combined with pituitary stalk sacrifice (HI+PSS) type (Figure 1C), hypothalamic injury combined with pituitary stalk sacrifice and pituitary gland injury (HI+PSS+PGI) type (Figure 1D), exclusive hypothalamic injury (HI) type (Figure 1E), exclusive pituitary gland injury (PGI) type (Figure 1F), and pituitary stalk sacrifice combined with pituitary gland injury (PSS+PGI) type (Figure 1G), were defined based on intraoperative videos. For the purpose of our study, we defined HI as the lack of the integrity of the hypothalamus, including mild HI, unilateral HI, and bilateral HI, which we have previously described elsewhere (15). Given that the pituitary stalk partly anatomically preserved may show function, only patients subjected to complete transection of the PS were included into the PSS group. Patients with intrasellar CP or supradiaphragmatic CP extending into the intrasellar were deemed as having PGI because of inevitable intraoperative manipulation of the pituitary gland. The confirmation of injury types was conducted by two independent neurosurgeons who were blinded for clinical outcomes. If they failed to reach consensus, a third neurosurgeon was assigned to make the final decision.

Presurgical and postsurgical endocrinological status was evaluated as described previously (16, 17). Partial hypopituitarism was defined as hormone deficiencies in one or two axes, and panhypopituitarism was defined as hormone deficiencies in three or more axes. Diabetes insipidus (DI) was evaluated preoperatively and postoperatively. Patients were diagnosed as having DI if they had polydipsia (excessive drinking; >3 l/day) and polyuria (excessive urination; >50 ml/kg body weight/24 h) combined with urine-specific gravity <1.005 and urine osmolality <300 mOsm/kg, as well as a positive response to desmopressin (15, 18). All the patients were recommended to receive imaging and clinical evaluation 3 months, 6 months, 12 months, and every 2 years subsequently after discharge. During follow-up, serum pituitary hormone levels were measured, and an endocrinologist analyzed the results. The protocol of hormone replacement was determined by the endocrinologist. Anterior pituitary dysfunction during follow-up was defined as replacement hormone required for at least one anterior pituitary hormone axis (14). For the purpose of our study, we focused on the endocrinological outcomes at 1 year after surgery and permanent DI representing posterior pituitary function was defined as at least 1 year of the need for desmopressin therapy after surgery (19).

All procedures were performed using an EEA. An image-guided neuronavigation system was employed for all procedures. All surgeries were performed by the chief surgeon (TH) and assisted by attending surgeons (BT, SHX, and LY). The surgical nuances of CP resection did not differ significantly from those described in the literature (20). After tumor resection, the closure was performed with a multilayered reconstruction, as described in previous literatures (21). Of note, the following surgical principles were strictly followed during CP resections. First, all procedures, especially in the process of tumor dissection, must be performed under direct visualization, avoiding pulling blindly. Second, carefully identifying and sharply combined with bluntly dissecting the cleavage plane between tumors and the HPA were warranted to achieve radical resection of CP and preserve the anatomical integrity of the structure of the HPA involved in the attachment to the greatest extent. When the stalk was infiltrated by tumors and extremely thin, the stalk would be sacrificed to achieve a radical resection. In addition, less aggressive resection was performed for patients with the extensive involvement of the hypothalamus or with adherence of a thin CP cyst capsule to critical vessels to avoid the higher risk of hypothalamic damage and fatal complications. Radiotherapy was routinely performed in CP patients undergoing STR 3 months after surgery.

All patients were randomly divided into a training group and a validation group in a 7:3 ratio. Continuous variables were reported as mean (± standard deviation) or median with an interquartile range (IQR), which was determined by the Shapiro–Wilk test. Categorical data were reported as counts and proportions in each group. The baseline characteristics between the two groups were compared via the chi-square test (Fisher’s exact test where appropriate) for categorical variables and two-tailed Student (Mann–Whitney U-test where appropriate) for continuous variables. Interrater reliability for classifying the injury type of the HPA was performed using Cohen’s kappa, indicating a good level of agreement (κ = 0.781). Univariate and stepwise multivariate logistic regression was used to determine independent risk factors associated with the need for anterior pituitary hormone replacement 1 year after surgery and permanent DI in the training cohort. First, we performed univariate logistic regression analysis using all the clinical variables associated with endocrinological deficiency after CP surgery based on our subject matter knowledge and the established risk factors identified by previous literatures (7, 10, 14). Then, a stepwise multivariate logistic regression analysis was performed using variables with p < 0.1 in the univariate logistic regression analysis. A nomogram was drawn based on the results of the multivariate logistic regression analysis. To evaluate the discrimination efficiency and calibration power of the nomogram, the area under the receiver operating characteristic (ROC) curve (AUC) and calibration curve with 1,000 bootstrap samples were employed in the training cohort, respectively, and validated in the validation cohort. Regarding AUC, a value >0.8 was thought to have a good discrimination according to the grading guidelines (22). At last, for clinical use of the model, the total scores were calculated based on the nomogram for each patient. The procedures of conducting the nomograms followed the recommendations of “seven steps for development and an ABCD for validation” proposed by Steyerberg et al. (23).

SPSS (version 25.0; IBM Corp., Armonk, New York) was used to perform statistical analyses and plot ROC curves, while GraphPad Prism version 8 (GraphPad Software, La Jolla, CA) was used to plot histograms. R software (version 4.0.3; http://www.Rproject.org) was used to plot nomograms and calibration curves, with the “rms” package used. A two-sided level of p < 0.05 was considered statistically significant unless indicated otherwise.

According to the inclusion and exclusion criteria, after excluding the patients undergoing reoperations, cases with perioperative mortality or loss to follow-up, and patients with endocrinological evaluation after surgery less than 1 year, 183 CP patients (104 men and 79 women) were enrolled into the final study cohort. The median follow-up duration of the entire cohort was 29.0 months (IQR 19–46 months). The included patients were randomly divided into the training (138 cases) and validation cohorts (45 cases).

Clinical data associated with risk of need for anterior pituitary hormone replacement 1 year after surgery and permanent DI are listed in Table 1. The baseline characteristics between these two cohorts were similar without significant difference. The number of permanent DI and needs for anterior pituitary hormone replacement 1 year after surgery was 71 (51.4%) and 94 (68.1%) in the training cohort, respectively, while 22 (48.9%) and 30 (66.7%) in the validation cohort, respectively.

Variables including presurgical, surgical, and postsurgical factors were selected to determine candidate risk factors via univariate logistic regression analysis. In this step, preop pituitary function (p = 0.001), tumor location (p = 0.012), and injury type (p < 0.001) were identified as potential risk factors and further included into stepwise multivariate logistic regression analysis to determine independent risk factors associated with the need for anterior pituitary hormone replacement 1 year after surgery (Table 2). On multivariate analysis, with results reported as odds ratio (OR, 95% CI), preop pituitary function (for partial hypopituitarism vs. normal, 4.184 [1.225–14.292]; for panhypopituitarism vs. normal, 13.742 [1.345–140.417]) and injury type (for HI vs. HPA-intact, 194.174 [21.311–1769.253]; for HI+PSS vs. HPA-intact, 89.443 [13.642–586.416]; for HI+PSS+PGI vs. HPA-intact, 68.111 [4.975–932.565]; for PSS vs. HPA-intact, 5.633 [0.753–42.133]; for PGI vs. HPA-intact, 159.790 [11.974–2132.357]; for PSS+PGI vs. HPA-intact, 31.921 [3.320–306.873]) were determined as independent risk factors and further used to form a nomogram (Table 3). Intriguingly, the difference in the risk of need for anterior pituitary hormone replacement 1 year after surgery between PSS and HPA-intact did not reach statistical significance, although stalk sacrifice had a higher risk (OR, 5.633; 95% CI, 0.753–42.133; p = 0.092).

Similarly, through univariate (Table 2) and stepwise multivariate analyses, preop DI (yes vs. no, 8.997 [2.580–31.374]) and the injury type of HPA (for HI vs. HPA-intact, 31.393 [6.319–155.964]; for HI+PSS vs. HPA-intact, 12.820 [3.138–52.380]; for HI+PSS+PGI vs. HPA-intact, 62.934 [5.795–683.512]; for PSS vs. HPA-intact, 1.580 [0.257–9.707]; for PGI vs. HPA-intact, 0.980 [0.164–5.848]; for PSS+PGI vs. HPA-intact, 2.399 [0.335–17.192]) were determined as independent risk factors associated with permanent DI and were used to form a nomogram. Note that when we compared PSS vs. HPA-intact, PGI vs. HPA-intact, and PSS+PGI vs. HPA-intact, the significant difference in the risk of permanent DI was not found (Table 3).

Further, we investigated the difference in endocrinological deficiency 1 year after surgery between the injury types of the HPA in the entire cohort (Figure 2). With regard to anterior pituitary dysfunction, the significant difference in the rate of need for anterior pituitary hormone replacement 1 year after surgery between injury types was found (p < 0.001). In addition, the rate of need for anterior pituitary hormone replacement 1 year after surgery was significantly lower in HPA-intact type and PSS type than HI type, HI+PSS type, HI+PSS+PGI type, PGI type, and PGI+PSS type (p < 0.05), whereas the difference between HPA-intact type and PSS type was not statistically significant (p > 0.05). Intriguingly, the difference in the rate of need for anterior pituitary hormone replacement 1 year after surgery between HI type, HI+PSS type, and HI+PSS+PGI type did not reach statistical significance (p > 0.05, Figure 2A). Regarding posterior pituitary dysfunction, there was a significant difference in the rate of permanent DI between injury types (p < 0.001). Moreover, the rate of permanent DI was significantly lower in HPA-intact type, PSS type, PGI type, and PGI+PSS type than HI type, HI+PSS type, and HI+PSS+PGI type (p < 0.05), whereas the difference between HPA-intact type, PSS type, PGI type, and PGI+PSS type was not statistically significant (p > 0.05). Likewise, the difference in the rate of permanent DI between HI type, HI+PSS type, and HI+PSS+PGI type did not reach statistical significance (p > 0.05, Figure 2B).

Nomograms based on the aforementioned independent risk factors to predict the probability of need for anterior pituitary hormone replacement 1 year after surgery and permanent DI are shown in Figure 3. The predicted probability of need for anterior pituitary hormone replacement 1 year after surgery and permanent DI for CP patients could be obtained, based on the sum of each variable score. Higher total scores calculated from the nomograms were associated with higher risk to suffer from endocrinological deficiency 1 year after surgery. For example, a CP patient with preoperative partial hypopituitarism, DI, and intraoperative pituitary stalk sacrifice would have a total of 110 scores (50 scores for partial hypopituitarism and 60 scores for pituitary stalk sacrifice) and a total of 92 scores (84 scores for preoperative DI and 18 scores for pituitary sacrifice), for a predicted need for anterior pituitary hormone replacement 1 year after surgery and permanent DI of 90.0% and 84.0%, respectively.

Further, ROC curves with AUCs and calibration curves were used to evaluate the performance of the nomograms. In the training cohort, for need for anterior pituitary hormone replacement 1 year after surgery and permanent DI predictions, the nomograms showed good discriminative abilities with AUCs of 0.921 and 0.885, respectively (Figures 4A, B). Calibration curves showed that the prediction curves were appropriately consistent with the ideal ones, which indicated that the nomograms had good calibration powers (Figures 4C, D). In addition, we validated the stability of the nomograms in the validation cohort. Similarly, good discriminative abilities with AUCs of 0.921 and 0.880 and moderate calibration powers of the nomograms were indicated by the ROC curves (Figures 5A, B) and calibration plots (Figures 5C, D), respectively.

First, due to the retrospective nature of the study, selection bias inevitably exists. Second, the present findings were based on limited data from a single center and on patients only undergoing EEA, it is necessary to validate them in multiple centers with a large sample size. Particularly, attention must be paid when interpreting our results that pituitary stalk infiltrated by tumors could be sacrificed, as only 15 patients underwent exclusive stalk sacrifice, warranting further investigation. Third, we defined endocrinological deficiency as the need for pituitary hormone replacement rather than objective measurement of pituitary hormone levels, which may underestimate the rate of pituitary deficiency, although it arguably gives a more clinically meaningful result (14). At last, the injury type of the HPA cannot be quantitatively defined, which may lead to bias when determining injury types between different neurosurgeons.