Serum Prolactin Level to Tumor Size Ratio as a Potential Parameter for Preoperative Differentiation of Prolactinomas from Non-Functional Pituitary Adenomas


 INTRODUCTIONPreoperative diagnosis of prolactinomas is critical because dopamine agonists have been regarded as a primary treatment modality. However, serum prolactin level alone is suboptimal for differentiating between prolactinomas and hyperprolactinemia-causing non-functioning pituitary adenomas (NFPAs). The authors investigated more effective parameter for differentiating prolactinomas and NFPAs by using the adenoma size. METHODS We performed a retrospective review of patients who underwent trans-sphenoidal surgery for pituitary lesions in a single institute between January 2015 and May 2021. Using the receiver operating curve (ROC) analyses, we compared performances of serum prolactin levels (PRL), a ratio of serum PRL levels to maximal tumor diameter (MD) (PRL/MD; PDR1), and MD squared (PRL/(MD)2; PDR2) in preoperative diagnosis of prolactinomas. RESULTS 223 patients with NFPAs (n=175) and prolactinomas (n=48) were included in the analysis. The prolactinoma group showed higher serum prolactin (258.6 ㎍/L) and smaller MD (16.6 mm) than those in the NFPA group (44.4 ㎍/L and 23.9 mm; both p-values < 0.001). Among diagnostic parameters, PDR2 exhibited the optimal diagnostic performance with the cutoff value of 0.83 [㎍/L]/mm2 (area under the curve [AUC] = 0.945), compared to the PDR1 (8.93 [㎍/L]/mm with AUC 0.938) and PRL (99.4 ㎍/L with AUC 0.910). PDR2 still maintained superior performance in the validation study than PDR1 and PRL (Accuracy of 94.8%, 91.8%, and 75.8%, respectively).CONCLUSIONS PDR2 provided the best performance of three parameters in preoperative discrimination of prolactinomas from NFPAs with hyperprolactinemia, and could contribute to select patients who benefit from medical treatment primarily.


Introduction
Prolactinoma is the most common type of pituitary adenoma (PA), accounting for 32-66% of all pituitary tumors requiring treatment. Current guid elines suggest the use of dopamine agonists (DAs) as a primary treatment for almost all spectra of prolactinomas from microadenomas to giant prolactinomas (maximal diameter [MD] > 40 mm) [1]. The preoperative diagnosis of PAs with hyperprolactinemia is a matter of debate. Hyperprolactinemia, the detection of serum prolactin (PRL) levels above the upper reference limit (commonly > 20 µg/L in men and > 25 µg/L in women) can have different physiological, pharmacological, and pathological causes [2][3][4]. PRL hypersecretion in PA is generally caused by a prolactinoma, which is the most common cause of hyperprolactinemia, or the "stalk section effect" of other PAs, in which the mechanical compression of the stalk blocks dopamine inhibition of lactotroph [2,[5][6][7].
Current endocrinology guidelines suggest several ranges of elevated PRL levels to distinguish between these two pathologies. Serum PRL levels > 500 µg/L are generally always indicative of prolactinomas [2,8]. Concerning lactotroph tumor cells secreting PRL, hyperprolactinemia of prolactinomas are known to be associated with tumor size [9][10][11]. While serum prolactin levels were usually at least 250 µg/L reaching 20,000 µg/L or more in macroprolactinoma, microprolactinomas with an MD less than 10 mm result in serum PRL levels of 100-200 µg/L [2,9].
Hyperprolactinemia of less than 100 µg/L is often associated with diagnostic uncertainty in the preoperative distinction between prolactinomas and non-functioning PAs (NFPAs). Up to 25% of microprolactinomas present with hyperprolactinemia < 100 µg/L [2,12]. Other studies have reported elevated PRL levels < 100 µg/L with a solid pituitary macroadenoma are highly suspicious of an NFPA [2,6]. These variations make serum PRL alone to be insu cient for discriminate prolactinoma from NFPA.
Thus the serum PRL levels reliable for discriminating prolactinomas from NPFAs remain unclear. The authors conducted a retrospective study to investigate the relationship between tumor size and serum PRL levels to investigate predictive factors distinguishing these two pathologies.

Patient enrollment
We performed a retrospective review of patients who underwent transsphenoidal surgery (TSS) for pituitary lesions between January 2015 and May 2021. This study was approved, and informed consent was waived by the Samsung Medical Center Institutional review board.
Among 997 consecutive patients with PAs who underwent TSS in this period, hyperprolactinemia, de ned as a serum PRL level > 25 µg/L, was con rmed at the time of initial diagnosis in 242 patients. We then identi ed the histopathology of these patients according to the World Health Organization (WHO) classi cation of pituitary tumors, using the version of 2004 and 2017 for patients who underwent TSS before 2017 and after 2017, respectively. Prolactinomas are de ned as lactotroph adenomas that mainly express PRL and related hormonal symptoms or signs such as galactorrhea-amenorrhea syndrome. NFPAs are de ned as PAs with a lack of clinical and biochemical evidence of adenohypophyseal hormone access. In the case of the diagnosis according to 2017 WHO classi cation, NFPAs included gonadotroph adenomas, silent corticotroph adenomas, silent adenomas of pit-1 derivation with no related hormonal excess and symptoms, and null cell adenomas. Finally, patients who were pathologically diagnosed as prolactinomas or NFPAs were included in the study. Other pathologies of PAs, including somatotroph adenoma, corticotroph adenoma, thyrotroph adenoma and plurihormonal adenoma were excluded.
Surgical indications of PAs were as follows: for prolactinomas, i) their adenomas were resistant to medical therapy at least 3-6 months after initiation, with persistent hormonal symptoms, elevated PRL levels, and constant or increased size of tumors, ii) patients are intolerable to the side effects of medical therapy, iii) patients who refused long term medications, due to preparation of pregnancy or preference for surgery, and iv) patients who were younger with PAs feasible for complete resection [8, 13,14]. For NFPAs, patients who presented with non-PRL-related signs and symptoms or sign related to the mass effect of PAs were indicated to undergo surgery. The presence of tumor apoplexy, manifesting as severe headaches and a sharp decrease in vision, indicated surgical treatment in both types of adenomas [15].
The preoperative serum PRL levels and maximal tumor diameters were obtained in all enrolled patients.
Especially in prolactinomas, we selected the initial serum PRL levels before medical treatment to minimize the effect of medication on volume change. The maximal diameter of the tumor was measured in axial, coronal and sagittal planes of diagnostic sellar magnetic resonance images (MRIs). The longest diameter of three measurements was de ned as "MD (mm)" in this study. The correlation between serum prolactin and tumor size was veri ed. Based on this relationship, the diagnostic value of the ratio of Comparisons between patients with prolactinoma and NFPA were performed in terms of general characteristics and relationships between tumor diameter and serum PRL levels. Pearson's correlation test was used to analyze the relationship between serum PRL level and tumor diameter according to each pathologic group. The diagnostic sensitivity, speci city, positive predictive value, negative predictive value, and the diagnostic accuracy of PRL, PDR1, and PDR2 were recorded. The accuracy was calculated as a conventional method used in model classi cation as below. Receiver operating characteristic (ROC) curves were used to investigate the optimal cutoff values of each parameter. The area under the curve (AUC) was compared using the student's t-test. We presented the 2X2 confusion matrix for validation results with Cohen's kappa coe cient and McNemar's test for the nal selection of the best cutoff model. P < 0.05 was considered statistically signi cant.

Patient demographic data
A total of 223 patients were included in the nal analyses, with 175 patients in the NFPA group and 48 patients in the prolactinoma group. The descriptive characteristics of the two groups are compared in Table 1. There was no signi cant difference in the proportion of sex between the NFPA and prolactinoma groups (25.7% versus 31.2%, respectively, p-value > 0.05). The median age of the NFPA group was higher than that of the prolactinoma group (43.0 years versus 30.0 years, respectively, p-value < 0.01). We further compared the serum PRL and MD parameters according to tumor size, particularly in the prolactinoma group (Table 2). Macroprolactinomas showed larger scales of both serum PRL levels and tumor MD than those of microprolactinomas, whereas the median value of both PDR1and PDR2 parameters was no signi cant difference between the two groups (p-value = 0.11). We conducted a correlation test between serum PRL levels and MD in each histological group. Pearson's correlation coe cients were used to estimate the linear relationship between the two variables. In the prolactinoma group, moderate strength of linear correlation between serum PRL and MD was con rmed in the positive direction (Pearson's r = 0.43, p = 0.002) < 0.05). In comparison, a low correlation was con rmed (Pearson's r = 0.17, p = 0.028) in the NFPA group ( Figure. 1).

Diagnostic performance of PRL, PDR1, and PDR2 in the study group
The ROC curve and the diagnostic power of PRL, PDR1, and PDR1 were calculated (Fig. 2). The optimal cutoff for indicating prolactinomas was determined using the Youden index, which maximizes the sum of sensitivity and speci city. The statistical difference between three AUCs were examined by the DeLong's test. The AUCs of PDR1 and PDR2 were signi cantly higher than PRL (PRL versus PDR1; p = 0.012 and PRL versus PDR2; p = 0.037). In contrast, there was no signi cant difference between the AUCs of PDR1 and PDR2 (p = 0.339). These results suggest that the ratio of serum prolactin to adenoma size showed better diagnostic powers than serum prolactin level alone in the study group.
Optimal parameter for preoperative diagnosis of prolactinomas: validation with the test group We conducted external validation with an independent cohort of 50 patients to select the optimal parameters among three cutoffs from ROC analyses. The validation group consisted of 13 patients with prolactinoma and 37 patients with NFPA. Table 1 described clinical characteristics of the cohort.
Cohen's Kappa coe cient, representing the degree of agreement between reference classi cation (clinical diagnosis in our study) and new classi er (PRL, PDR1, and PDR2), also revealed strong consistency in PDR2 (k = 0.896). PDR1 showed a moderate-to-strong degree kappa coe cient of 0.729, while kappa of PRL alone was low (k = 0.384). To further evaluate whether the statistical difference exists in classi cation between reference and new predictor, we used McNemar's test -PRL; p = 0.001, PDR1; p = 0.041 and PDR2; p = 1.000. As a p-value more signi cant than 0.05 in McNemar's test suggests no signi cant difference between the two classi ers, the prediction of prolactinomas with PDR2 was statistically consistent with the reference. These results suggested that PDR2 is the optimal model among three parameters for differentiating prolactinomas from NFPAs.

Discussion
This study investigated the predictive value of PRL and PDR for preoperative differentiation of prolactinomas and NFPAs. Distinguishing these two pathologies is critical given the satisfactory response to DAs in prolactinomas and the need for surgical resection in large NFPAs. The European Endocrine Society suggested serum PRL levels > 250 µg/L in macroadenomas (diameter > 1 cm) as a clinical diagnostic threshold for prolactinomas [8]. In our data, this threshold was insu cient for the diagnosis of prolactinomas. 24 (50%) of 48 patients with prolactinoma exhibited mild elevation of serum PRL levels < 250 µg/L at the initial diagnosis, with 9 patients of microadenoma and 15 of macroadenoma. Five (10.4%) of 48 patients presented with an even low degree of hyperprolactinemia with serum PRL levels < 100 µg/L.
NFPA and other sellar masses (growth hormone, adrenocorticotropic hormone, or thyroid-stimulating hormone; craniopharyngiomas; hypophysitis; etc.) are known to be typically associated with mild hyperprolactinemia < 100µg/L[8, 12,19,20]. These results were consistent with our results, as 142 (81.1%) of 175 NFPAs in this study displayed a serum PRL level of less than 100 µg/L. We found moderate-to-high levels of hyperprolactinemia in some patients with NFPA, as 27 (15.4%) showed serum PRL levels between 100 and 250 µg/L. Serum PRL levels of outliers, approximately > 1,000 µg/L, were checked in three patients in the NPFA group, which all were macroadenoma with tumor diameter larger than 3 cm. Thus the maximal PRL levels found in non-functional macroadenomas are still a matter of debate.
Previous studies reported that hormonal symptoms were much more prevalent than mass effects in prolactinomas and vice versa in NFPAs [9,21]. We found similar ndings that hyperprolactinemic symptoms such as amenorrhea, galactorrhea was more common in prolactinomas (62.6% in prolactinomas vs 25.1% in NFPAs, p < 0.001) and tumor mass effects with headache or visual disturbance were more prevalent in NFPAs (50.9% in NFPAs vs. 29.2% in prolactinomas, p < 0.05). These different clinical manifestations might help differentiate prolactinomas and NFPAs.
Tumor size and serum PRL levels displayed different relationships in our study. Prolactinomas showed a moderate linear correlation between serum PRL and tumor MD in the positive direction (Pearson's r = 0.43, p-value = 0.002), while NFPAs exhibited a weak correlation between the two parameters (Pearson's r = 0.17, p-value = 0.028). These results were comparable to other studies [1,8,9,22], which suggest that lactotroph tissue is more contributable to the hyperprolactinemia than the stalk section effect due to the tumor mass size [23][24][25].
Based on these results, several attempts were made to incorporate adenoma size into differentiating prolactinomas and hyperprolactinemia-causing NPFAs. Burke [10]. Wright et al. used the ratio of serum PRL to tumor volume (PRL/V) for diagnosis of prolactinomas (n = 21) from NFPAs (n = 58), suggesting 21.62 [ng/mL] / as the cut off value with a sensitivity of 100% and speci city of 82.76%. However, the statistical difference in diagnostic performance between the ROCs of PRL/V and serum PRL level alone was not described [11]. Further investigation is needed due to the smallsized sample and heterogeneous pathologies in the control group.
In this study, we examined and validated the diagnostic performance of the novel parameters, the ratio of PRL to MD (PDR1) and MD squared (PDR2), to that of PRL alone. The optimal cutoff values were 99.42 µg/L for PRL (AUC = 0.910 and accuracy 82.9%), 8.93 [µg/L]/mm for PDR1 (AUC = 0.938, accuracy = 89.7%) and 0.83 [µg/L]/mm 2 for PDR2 (AUC = 0.945 and accuracy 93.3%). Both PDR models had superior outcomes in the ROC analyses than PRL, and in the validation study, the PDR2 model was the best classi er with statistical signi cance. These results suggested that considering the prolactin-productivity per adenoma size may improve the preoperative prediction of prolactinomas.
The limitation of this study includes the bias from its retrospective nature and sample sizes susceptible to outlier effects. One of the major limitations is selection bias; data were obtained only from surgical cases, causing relatively low "prevalence" of prolactinomas compared to NFPAs in our clinical setting. In the future, multicenter and prospective clinical studies are required to improve the accuracy and further elucidate the role of PDR in the differential diagnosis of prolactinomas from hyperprolactinemia-causing other pituitary pathologies.
In conclusion, our study demonstrated the effectiveness of serum PRL to tumor size ratio as a potential parameter for preoperative differentiation of prolactinomas and NFPAs. Based on the positive correlation between serum PRL and tumor MD in prolactinomas, contrary to the weak relationship observed in NFPAs, we examined and validated the diagnostic value of the PDR parameters compared to PRL alone.
The optimal thresholds of the PRL to MD squared ratio may contribute to preoperative diagnosis of prolactinomas from other conditions of PAs, hence improving a treatment strategy whether administration of DA agonist or surgical resection should be recommended.

Declarations
Funding: Not applicable.
Con icts of interest/Competing interests: The authors report no con ict of interest concerning the materials or methods used in this study or the ndings speci ed in this paper.
Availability of data and material: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.  Figure 1 Correlation between serum prolactin level and tumor diameter according to histologic groups.

Figure 2
Comparison of the ROC curves and diagnostic performance of serum prolactin level (PRL) and a ratio of serum prolactin to adenoma maximal diameter (PRL/MD, PDR1) and to diameter squared (PRL/(MD)2, PDR2) for differentiating prolactinomas from non-functioning pituitary adenomas. Figure 3