Method Matters: Does Measuring Hormones with LCMS or Immunoassay Produce Different Relationships with Psychopathology?

• In Event: Measurement Issues in Developmental Research

Abstract

Methodological comparisons of techniques for measuring levels of steroid hormones have found divergence between immunoassay results and mass spectrometry results, showing that results obtained with immunoassay tend to overestimate low hormone values relative to mass spectrometry (Schultheiss et al., 2016). Additionally, it has been demonstrated that immunoassay tends to underestimate gender differences in hormone levels in testosterone due to inflation of female testosterone levels which are naturally low (Welker et al., 2016). Despite this growing body of research, very few studies have examined how differences in analytical techniques may be affecting relationships between steroid hormone levels and demographic, behavioral, or psychological measures (Bae et al., 2016). We will present the results of a pre-registered analysis of an existing dataset in which three days of salivary cortisol and testosterone were measured using both Chemiluminescent immunoassay (IA) and liquid chromatography dual mass spectrometry (LC/MS-MS) in roughly 185 9th graders from a single high school in Texas. These data included samples collected before and after administration of a behavioral intervention. We completed a preliminary analysis using all three days of hormone values in order to confirm that hormone values obtained using IA did differ from those obtained using LC/MS-MS in our sample. We found that measurement method was associated with systematic differences in hormone levels, with IA producing inflated estimates of testosterone and cortisol, especially at low levels of hormones and especially in testosterone measured in female participants. Additionally, we used pre-intervention measures of hormone levels to examine whether differences in hormone levels obtained using these two techniques were associated with different relationships between hormone levels and psychological measures (measured by self-report scales including the Children’s Depression Inventory, the Perceived Social Stress Scale, the UCLA Loneliness Scale, and the Anxious Avoidant and Negative Self Evaluation subscales of the Social Anxiety Scale for Adolescents). The results of our analyses, which employed generalized additive models in order to explore parametric and non-parametric hormone-psychology relationships, demonstrated differential linear and non-linear relationships between hormones and psychological measure scores. These differing relationships have important implications for how we consider findings in which hormones measured using immunoassays have been related to psychological measures and suggests the broad impact of the techniques we use to measure hormone levels in studies examining hormone-psychopathology relationships.

Authors

• Julia E Chafkin, University of Texas at Austin
  Presenting Author

• Joseph O'Brien, University of Texas at Austin
  Non-Presenting Author

• Robert Alan Josephs, University of Texas at Austin
  Non-Presenting Author

• Fortunato Nicholas Medrano, University of Texas at Austin
  Non-Presenting Author

• David Scott Yeager, University of Texas at Austin
  Non-Presenting Author