Comparing Neuroendocrine Recovery Between Surgical and Conservative Management in Pituitary Apoplexy Patients: A Propensity Score-Matched Analysis

Background: Pituitary apoplexy is a rare and potentially life-threatening clinical syndrome. Patients may present with severe neuro-ophthalmologic or endocrine symptoms. Current evidence is unclear whether conservative or surgical management leads to the best neuroendocrine outcomes. This study aimed to compare neuroendocrine outcomes between surgical and conservative treatments in a single center. Methods: Cases of patients with pituitary apoplexy who received transsphenoidal surgery or conservative management in Songklanagarind Hospital between January 1, 2005 and December 31, 2022 were retrospectively reviewed. A propensity score matching method was used to adjust bias from treatment selection (surgery or conservative treatment). Differences in visual acuity, visual �eld, cranial nerve, and endocrine outcomes between the surgical and conservative treatment groups were analyzed using logistic regression analysis. Results: This study included 127 patients, with 98 and 29 patients in the surgical and the conservative treatment group, respectively. The optimal matching method was used for propensity score matching. Compared to the conservative group, the surgically treated patients had a signi�cantly higher rate of visual �eld recovery (odds ratio (OR): 12.89, P = 0.007). However, there were no statistical differences in the recovery rate of preoperative visual acuity, cranial nerve, and endocrine de�cits between the groups. Conclusions: Transsphenoidal surgery was associated with a higher rate of visual �eld recovery when compared to the conservative treatment for pituitary apoplexy patients. Careful selection of appropriate treatment based on the patient’s presentation and neuroendocrine status will result in the best outcomes while avoiding unnecessary surgical intervention.


INTRODUCTION
Pituitary apoplexy is a rare clinical syndrome resulting from rapid expansion within the sellar turcica due to either hemorrhage or infarction [4].This clinical syndrome is characterized by headache, visual de cits, ophthalmoplegia, and alteration in mental status [2].
Depending on the patient's symptoms and neuro-ophthalmologic and endocrine status, surgical or conservative treatment is chosen.Currently, there is no clear surgical indication for these patients.The outcomes of surgical and conservative treatment strategies for patients with pituitary apoplexy have been reported in several studies [1,3,5,7,[10][11][12]14].However, there is no clear evidence whether surgical or conservative treatment will provide the best visual, cranial nerve (CN), and hormonal outcomes.The interpretation of previous results is limited by small study size, difference in patient characteristics between the two treatment groups, and bias from treatment selection.
This study aimed to compare neuroendocrine recovery outcomes between surgically and conservatively treated patients by using the propensity score matching method to adjust for bias from treatment selection.Four major outcomes were evaluated in this study: visual acuity (VA), visual eld (VF), CN, and endocrine recovery from preoperative de cits.

Study Design and Patients
This study was performed in line with the principles of the Declaration of Helsinki.Approval was granted by the Ethics Committee of Prince of Songkla University (17 February 2023/No.REC.66-038-10-4).We retrospectively reviewed the records of consecutive patients with pituitary apoplexy who received conservative treatment or underwent transsphenoidal surgery for tumor removal at Songklanagarind Hospital, a tertiary care center and medical school in southern Thailand, between 1st January 2005 and 31st December 2022.Participants were recruited from our database of patients with pituitary tumors who received treatment in Songklanagarind Hospital.Patients who had radiographic evidence of pituitary apoplexy together with neuro-endocrine symptoms (VA and/or VF defect, CN III-VI palsy, or hypopituitarism) were included in this study.Up until August 2014, a microsurgical approach was used for transsphenoidal surgery in our hospital.After that date, we only used an endoscopic approach owing to the advantage of wider operative exposure offered by this technique.Each patient was evaluated and treated by the endocrinologist, ophthalmologist, and neurosurgeon.All transsphenoidal operations were performed by either or both authors (T.O. and C.T.).Intraoperative computed tomography-guided neuronavigation was used in all endoscopic endonasal operations.

Measurements
Medical records were reviewed and data regarding patient demographics, comorbidities, preoperative use of antithrombotic medication, clinical presentation, VA, VF, CN III-VI de cits, pituitary hormonal status before and after treatment, imaging studies, adjuvant treatment, complications, and duration of follow-up were collected and analyzed.Recovery of VA after treatment was de ned as a logMAR improvement of at least 0.1 between pre-and post-treatment status measured at the last follow-up evaluation.Recovery of VF after treatment was de ned as any improvement of visual eld defects following treatment.Recovery from a CN de cit and hypopituitarism (hypoadrenalism, hypothyroidism, or hypogonadism) was evaluated at the last follow-up evaluation (at least 6 months after surgery) and documented.
The superior, lateral, and inferior extensions of the lesion were evaluated using the Knosp and Wilson-Hardy classi cation systems [9,16,17].Lateral extensions were evaluated using the Knosp classi cation system and then classi ed into two groups: grades 0-2 and 3-4.Superior extension was evaluated using the Wilson-Hardy suprasellar extension stage and then classi ed into two groups: stages 0-A and stages B-C.Inferior extension was evaluated using the Wilson-Hardy invasion grading system and then classi ed into no (grades 1-2) or present (grades 3-4) sphenoid extension.The height of any suprasellar extensions was measured in millimeters to quantify the suprasellar extension of the lesion.An imaginary line was drawn from the tuberculum sellae to the dorsum sellae, and the suprasellar extension was measured by using a perpendicular line from the imaginary line to the superior part of the lesion [13].
The visual impairment score (VIS) developed by Fahlbusch and Schott [6] was used to quantify the amount of visual impairment prior to treatment.The degrees of VA and VF defects in each patient were translated into scores using speci c tables.Afterward, the sum of the VA and VF defect scores in each patient was calculated, providing a VIS ranging from 0 (best) to 100 (worst) [6].

Statistical Analysis
Sex, age, comorbid conditions (diabetes, hypertension, underlying bleeding tendency), current antithrombotic medication, clinical presentation, pre-treatment VA, VF, CN, and pituitary hormonal status, imaging studies, and adjuvant treatments were analyzed and compared between surgically and conservatively treated patients.Logistic regression analysis was used to identify factors associated with treatment selection (surgical or conservative treatment).Then, the propensity score was calculated based on these factors.The patient populations of the two groups were matched using various propensity score matching techniques.The matching technique that could optimally balance the covariates was chosen for comparing the outcomes between surgical and conservative treatment.
The population after matching was then further analyzed.Baseline characteristics of patients were analyzed and compared between the two treatment groups.The neuroendocrine outcomes (VA, VF, CN, and pituitary hormone recovery) were compared between the groups.Logistic regression analysis was used to evaluate the association between treatment approach (surgery or conservative) and each neuroendocrine outcome.The multivariate analysis model with the lowest Akaike information criterion value was chosen to identify the factors associated with each neuroendocrine outcome.The statistical analyses were performed with the R Program version 4.3.2(R Foundation for Statistical Computing, Vienna, Austria).A P value < 0.05 was considered statistically signi cant.

Baseline Characteristics before Matching
The study included 127 patients.Ninety-eight patients underwent transsphenoidal surgery, while 29 patients received conservative treatment.Table 1 shows the baseline characteristics of the patients before propensity score matching.Thirteen (45%) and 47 (48%) patients in the conservative and the surgery group, respectively, were male.The mean age of the patients in the conservative group was signi cantly older than that in the surgery group (58.7 vs. 51.6 years, P = 0.019).There were no differences in underlying comorbidities between the groups.Seven (24%) and 10 (10%) patients in the conservative and the surgery group, respectively, had been using antithrombotic medication before the onset of pituitary apoplexy.Patients in the conservative group presented with headache more frequently (P < 0.001), while patients in the surgery group had a higher incidence of preoperative VF de cit on both sides (P = 0.005).The lower incidence of VF de cit (compared to VA de cit) might have resulted from an inability to accurately measure VF in patients with poor VA.Preoperative VIS and cranial nerve de cits showed no signi cant differences between the two groups.Patients in the surgery group had the higher incidence of preoperative hypothyroidism (52% vs. 28%, P = 0.035) and hypogonadism (63% vs. 31%, P = 0.006).The preoperative median serum prolactin level was signi cantly higher in the surgery group (22.4 vs. 10 ng/mL, P = 0.019).The median tumor size was larger in the surgery group (33.5 vs 18.1 mm, P < 0.001).Classi cation of the patients using the Knosp, Wilson-Hardy suprasellar extension, and Wilson-Hardy invasion classi cation systems showed signi cantly higher grades in the surgery group than in the conservative group (P < 0.001, P < 0.001, and P = 0.014, respectively).The median suprasellar extension of the lesion was higher in the surgery group (16.7 vs. 3.9 mm, P < 0.001).The incidence of tumor recurrence or enlargement following treatment showed no difference between the groups.Patients in the surgery group received adjuvant treatments more often than those in the conservative group (P < .001).

Baseline Characteristics after Matching
Logistic regression analysis was used to identify factors associated with surgical intervention in all patients.Then, these factors (headache symptom, VIS, Knosp classi cation grade, and suprasellar extension grade) were used to calculate the propensity score (Fig. 1).Subsequently, various propensity score-matching methods, including the nearest neighbor matching with the ratios 1:1 and 2:1, Caliper matching, optimal matching, optimal full matching, and Mahalanobis metric matching were used for matching the populations of both groups.The optimal matching technique was chosen as the optimal method to balance the covariates before further analysis.
Table 2 shows the baseline characteristics of the patients after propensity score matching with the optimal matching technique.There were 28 patients in each group, and the mean age, headache symptom, preoperative VF de cit on both sides, median serum prolactin level, and imaging classi cation grade showed no signi cant differences between the groups after matching.However, patients in the surgery group still had a signi cantly higher incidence of preoperative hypothyroidism and hypogonadism, larger median tumor size, and higher suprasellar extension (P = 0.03, P = 0.011, P < 0.001, and P < 0.001, respectively).

Neuroendocrine Outcomes
Table 3 shows the neuroendocrine recovery after conservative and surgical treatments in patients who had preoperative de cits.VA recovery showed no signi cant difference between the surgery and conservative groups.Nevertheless, patients in the surgery group had signi cantly better VF recovery in at least one eye after treatment (85% vs. 50%, P = 0.034).Recovery of CN III palsy showed no difference between the groups.The incidences of preoperative CN IV, V, and VI palsy were too low to allow comparison of the outcomes between the groups.Recovery from hypoadrenalism, hypothyroidism, and hypogonadism showed no difference between the groups.

Factors Associated with Neuroendocrine Recovery
Logistic regression analysis was identify factors associated with VA recovery in at least one eye after treatment.Table 4 shows the multivariate analysis of factors associated with recovery from preoperative VA de cits.Older age was signi cantly associated with no recovery from a preoperative VA de cit (P = 0.029).Surgery was not associated with better VA recovery (P = 0.529).Table 5 shows the multivariate analysis of factors associated with recovery from a preoperative VF defect.Absence of underlying hypertension and surgical treatment were signi cantly associated with better VF recovery (P = 0.013 and P = 0.007, respectively).Results from the logistic regression analysis showed that surgical treatment was not associated with better recovery from preoperative hypoadrenalism, hypothyroidism, and hypogonadism.
Owing to the low incidences of preoperative CN III-VI de cits and the very high incidence of CN recovery following treatment, the association between surgical treatment and CN recovery outcome could not be analyzed.

DISCUSSION
Before propensity score matching, patients in the conservative group were older and presented more frequently with headache, while patients in the surgery group had a higher incidence of preoperative VF de cit, hypopituitarism, larger tumor size, and greater tumor extension in all directions.We used propensity score matching analysis to balance these covariates between the conservative and the surgery group.Following propensity score matching with the optimal matching method, the mean age, headache symptom, and preoperative VF de cit were not signi cantly different between the two groups.However, patients in the surgery group still had a signi cantly higher incidence of preoperative hypothyroidism and hypogonadism, larger median tumor size, and higher suprasellar extension.Therefore, we used a multivariate analysis to control the remaining possible confounding factors and evaluate the association between surgical (vs.conservative) treatment and neuroendocrine recovery.
Our study found that compared with conservative treatment, surgical treatment was signi cantly associated with better improvement of preoperative visual eld de cit in at least one eye.Nevertheless, surgical treatment was not associated with better recovery of VA, CN III, and hypopituitarism de cit.
Previous studies found no signi cant differences in VA and VF recovery between surgically and conservatively treated patients [1,3,5,[10][11][12]14].A meta-analysis study by Goshtasbi et al. [8] also found no difference in visual outcomes between surgically and conservatively treated patients.However, a meta-analysis study by Tu et al. [15] found that surgically treated patients had a signi cantly higher rate of recovery of visual eld (P < 0.05), while there was no signi cant difference in VA recovery between the surgical and the conservative group.
In all of these studies [1,3,5,[10][11][12]14], compared to the conservative treatment group, the patients in the surgery group had more severe visual symptoms.Furthermore, the tumor in the surgical group was larger and showed greater extension than that in the conservative treatment group [1,3,5,[10][11][12]14].
This potential bias might interfere with outcomes comparison between the surgery and conservative groups.In our study, we used propensity score-matched analysis to balance the covariates and then used a multivariate analysis to assess the association between surgical (vs.conservative) treatment and various neuroendocrine outcomes.
Our study found no difference in CN III recovery between the surgery and the conservative group.Most previous studies found no difference in CN recovery between the surgery and the conservative group [1,3,5,8,10,12].A study by Shepard et al. [11] found better cranial neuropathy improvement in the conservatively managed patients (p < 0.01).On the contrary, a meta-analysis study by Tu et al. [15] found a signi cantly better recovery of ocular palsy in the surgically treated patients (P < 0.05).
Recovery from hypoadrenalism, hypothyroidism, and hypogonadism showed no differences between the groups in this study.Likewise, studies by Giritharan et al. [7], Singh et al. [12], and Bujawansa et al. [3] found no differences in endocrine outcomes between the surgery and the conservative group.A metaanalysis by Goshtasbi et al.
[8] and Tu et al. [15] also found no differences in improvements of endocrine dysfunction between surgically and conservatively treated patients.A study by Teixeira et al. [14] however found better endocrine outcomes in the surgical group (P = 0.027), while a study by Marx et al. [10] found more endocrine de cits at 1 year in the surgical group (p = 0.029).
This study also found that older age and underlying hypertension were signi cantly associated with poor recovery from a preoperative VA and VF de cit, respectively.These ndings were consistent with our ndings in a previous study [13].
There are some limitations in this study.First, the retrospective study design potentially introduces bias and confounding factors.However, we used a propensity score-matched analysis to balance the covariates between the surgery and the conservative group.We also used a multivariate analysis to determine the association between surgery (vs.conservative treatment) and various neuroendocrine outcomes in the matched population.Second, we included patients with pituitary apoplexy in our institute between 2005 and 2022, and the evolution of medical care and operative treatment during this period might affect the treatment outcomes.Third, the baseline characteristics of patients in the conservative and surgery groups were different.Even after propensity score matching, some confounding factors remained.Therefore, we used a multivariate analysis to clarify the effect of surgery (vs.conservative treatment) on various neuroendocrine outcomes.
In terms of strengths, the population in this study was relatively large, even after propensity score matching, compared with that in previous studies.Second, the propensity score-matched analysis potentially eliminates the bias from treatment selection (surgery or conservative treatment) for each patient with pituitary apoplexy in our study.Third, the better VF recovery in the surgery group, compared with that in the conservative treatment group, has not been identi ed in the previous studies, excepted the meta-analysis study by Tu et al. [15].This nding supports the bene t of surgery in patients with pituitary apoplexy who present with VF de cit.

CONCLUSIONS
Patients with pituitary apoplexy usually present with headache or visual impairment.Most of these patients recovered from visual de cit and CN palsy but still had endocrine de cit following treatment.Our

1
Abbreviations adjusted odds ratio

Table 1
Baseline Characteristics of Patients Before Propensity Score Matching Data are presented as n (%) unless indicated otherwise.IQR, interquartile range; SD, standard deviation; VA, visual acuity; VF, visual eld.

Table 2
Baseline Characteristics of Patients After Propensity Score Matching with Optimal Matching Technique IQR, interquartile range; SD, standard deviation; VA, visual acuity; VF, visual eld.

Table 3
Neuroendocrine Recovery After Conservative and Surgical Treatment in Patients with Preoperative De cits VA, visual acuity; VF, visual eld; CN, cranial nerve.

Table 4
Multivariate Analysis of Factors Associated with Visual Acuity Recovery in One or Both

Table 5
Multivariate Analysis of Factors Associated with Visual Field Recovery in One or Both Eyes After Treatment