Materials and Methods
Patient Selection and Preoperative Assessment
This study was approved by the USC Institutional Review Board (IRB HS-11-00702). Patients surgically treated by authors M.W. and G.Z. for prolactinomas at the USC Pituitary Center from 1995 to 2020 were identified from a prospectively maintained RedCap database and retrospectively reviewed. All patients had prolactinomas confirmed by histopathological assessment. One patient did not have pathology available because the tumor was suspected to be suctioned before a sample could be obtained; however, it did have preoperative prolactin greater than 200 ng/mL that was reduced with DA therapy and exhibited biochemical remission postoperatively. Resection was offered as second-line treatment for patients who failed DA therapy or had persistent significant symptoms secondary to mass effect. In some cases, the preoperative diagnosis was indeterminate and the diagnosis of prolactinoma was made following histopathological analysis. In rare cases, resection was offered as a first-line treatment if patients expressed a strong preference for surgery and a clinical decision was made that the tumor could be completely resected and potentially cured. All patients underwent a thorough preoperative workup including a complete history, physical examination, endocrine laboratory testing, ophthalmological exam when appropriate, and magnetic resonance imaging (MRI) to assess the size of the tumor and determine the extent of invasion when present. Extrasellar extension was defined on MRI as tumor growth into the cavernous sinus, frontal lobe, clivus, sphenoid sinus, or suprasellar region. Degree of invasion was determined and scored by an independent neuroradiologist for this analysis. The minimum follow-up time for inclusion was 3 months.
All patients underwent a standard direct microscopic or endoscopic endonasal transsphenoidal approach based on surgeon preference. Intraoperative cerebrospinal fluid leaks were repaired using autologous adipose grafts for packing of the sella and/or fascial reconstruction of the sellar dura using a 2-layer fascial apposition technique.
Follow-up and Outcome Measurements
In the immediate postoperative period, patients were monitored for diabetes insipidus (DI), hyponatremia secondary to syndrome of inappropriate antidiuretic hormone secretion, and other electrolyte imbalances. Prolactin and electrolyte levels were drawn on postoperative days (POD) 1 and 2, and serum sodium levels were checked on POD7. After discharge, patients were followed by clinical assessment, MRI, measurement of prolactin levels, and visual field testing at intervals of 1.5, 3, and 6 months for the first year after surgery. Patients with residual tumor were evaluated for adjuvant medical therapy and radiation therapy or stereotactic radiosurgery (SRS), which was offered at the discretion of the treating radiation oncologist and neurosurgeon based on residual tumor size and location. Degree of resection was determined by an independent neuroradiologist.
Patients were divided into 4 groups for the purposes of analysis: surgical cure, surgical noncure, biochemical control, and nonbiochemical control. Surgical cure was defined as normalization of prolactin levels following surgery without the need for postoperative DA, whereas biochemical control was defined as normalization of prolactin levels following surgery, either with or without the use of postoperative DA. These determinations were based on laboratory data and medical history at the time of the patient's last known clinical visit. These categorizations are consistent with those used in prior literature and allow for overlap between outcome groups.
Whenever possible, postoperative MRI results and serum prolactin levels were used to determine prolactinoma recurrence (elevation of prolactin above normal) and progression (tumor growth on MRI). When postoperative MRI results were not available within the electronic medical record (EMR), recurrence/progression was determined by review of evaluations made by board-certified endocrinologists, adhering to these definitions.
Descriptive statistics were used to summarize patient characteristics. Categorical variables were analyzed with chi-square analysis. Continuous variables were analyzed with independent t test (for normally distributed data) or Mann-Whitney U test (if distribution was not normal). SPSS 25.0 statistical software (IBM Corp) was used to perform the statistical analysis and generate receiver operating characteristic curves. P values less than .05 were considered to be statistically significant.
Prolactin Level Analysis
Preoperative prolactin levels were defined as the most recent prolactin level of patients while they were undergoing DA therapy prior to surgery (n = 47). Individuals who did not undergo DA therapy prior to surgery were excluded from analysis of preoperative prolactin level calculations because they had disproportionally elevated prolactin levels that were not predictive of the surgical response to treated prolactinomas relative to the other patients in our study (n = 9). For these patients, it is possible that DA therapy was either discontinued or not recorded in the EMR. POD1 prolactin was measured as a morning laboratory draw the first day after surgical resection using the Elecys Prolactin II immunoassay (Roche Diagnostics).
To aid in analysis of our results, a systematic review of studies reporting the surgical management of prolactinomas through the PubMed and Google Scholar databases was conducted. Articles were included in the analysis if all cases in the series were surgically managed either as a primary mode of treatment, along with DA therapy, radiotherapy, or a combination of these treatment modalities. Series of patients managed solely by DA therapy or radiosurgery were excluded. Case reports, mixed-pituitary tumors, systematic and literature reviews, and cases looking at special populations, such as articles focused on pregnancy, pediatric adenomas, or multiple endocrine neoplasia type 1 syndrome, patients were excluded. For practicality, all articles not in English and series that did not examine predictors of hormonal remission were also excluded. A total of 160 studies were identified through PubMed and 17 additional studies were identified through Google Scholar and from citations of included papers, for a total of 177 initial records after duplicates were removed (Figure 1). Of these, 38 full-text articles were assessed for eligibility, and 19 studies published from 1992 to 2020 were included in the analysis.
J Endo Soc. 2021;5(10) © 2021 Endocrine Society