Banca de DEFESA: Alan Carvalho Dias

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : Alan Carvalho Dias
DATE: 06/07/2023
TIME: 09:00
LOCAL: Sala de Reuniões 1 da Faculdade de Ciências da Saúde
TITLE:

EVALUATION OF MACHINE LEARNING TOOLS TO RELATE PROLACTIN CONCENTRATION AND METABOLIC HOMEOSTASIS

KEY WORDS:

hypoprolactinemia; insulin resistance; metabolic syndrome; hyperprolactinemia, machine learning

PAGES: 100
BIG AREA: Ciências da Saúde
AREA: Medicina
SUMMARY:

Introduction: Recently, a new "metabolic classification" was proposed for prolactin, and based on it, hypoprolactinemia and hyperprolactinemia were defined as results approximately < 7 and > 100 ng/mL, respectively. Various metabolic diseases have been associated with this range of results, however, the determination of the cut-off points that influence metabolism seem uncertain. Objectives: To investigate the possible correlation between prolactin concentration and glucose and lipid metabolism tests. To establish a "gray zone" that represents the inflection points, associating the results of the glycemic and lipid metabolism tests with prolactin. As part of the study, a machine learning tool was created using the R language to perform these analyses in an automated way. Methods: A cross-sectional study was conducted with 65,795 laboratory results from adult patients of both sexes, collected in the first half of 2018 at the Sabin Laboratory. The data, extracted and anonymized from the Laboratory Information System, included lipid parameters (total cholesterol, HDL-c, LDL-c, triglycerides) and biochemical/hormonal parameters (glucose, prolactin, insulin, HOMA-IR). Individual data were stratified into 106 partitions, based on the average concentrations of prolactin and lipid and glycemic metabolism analytes of each individual. These averages were used in three analyses: (1) in the comparison of the results of the glycemic and lipid metabolism tests in each range of the new metabolic classification of prolactin, (2) in the estimation of the inflection point through a machine learning model and the "gray zone" that represents the equivalent inflection points, associating the results of the glycemic and lipid metabolism tests with prolactin, and (3) in the prediction of the average results of these glycemic and lipid metabolism tests from the estimated inflection point. In addition, individual prolactin results were used to compare prolactin concentrations between sexes. Results: The analyzed population was aged between 21 and 75 years, composed of 51,366 women and 14,429 men. Hyperprolactinemia (prolactin concentrations above 25 ng/mL) was identified in 4,004 of the 51,366 women (7.79%) and in 208 of the 14,429 men (1.41%). Prolactin results were equivalent between sexes. The average results of the glycemic and lipid metabolism tests were significantly higher in the range of average prolactin results < 7 ng/mL. In the HomeoFIT-PRL range, corresponding to the range of the median and the distribution of the average results of the glycemic and lipid metabolism tests are predominantly lower compared to the other ranges of average prolactin results. Below the "gray zone" represented by the average prolactin results between 9.58 and 12.87 ng/mL, there may be impairments in glycemic metabolism, while for lipid metabolism, the estimated "gray zone" was 13.81 to 18.73 ng/mL. Conclusion: Our research confirms the correlation between variations in average prolactin concentrations and glycemic and lipid metabolism tests in humans, suggesting a mediating role of prolactin in the pathogenesis of impaired glucose and lipid metabolism. However, the cross-sectional nature of the study limits causal inference. Prospective studies with large data volumes and appropriate statistical approaches are needed to clarify the causeeffect relationship. The historical average of prolactin concentrations may offer more precise insights for understanding metabolic homeostasis. 

BANKING MEMBERS:
Interna - 2329402 - ANGELICA AMORIM AMATO
Presidente - 2202820 - LUCIANA ANSANELI NAVES
Externo ao Programa - 4123148 - MARCELO PALMEIRA RODRIGUES
Externo à Instituição - RAFAEL HENRIQUES JACOMO - SABIN