Harmonization of Neuroticism and Extraversion phenotypes across inventories and cohorts in the Genetics of Personality Consortium: an application of Item Response Theory

Stéphanie Martine Van Den Berg ,Alison Pattie,Ina Giegling,Paul T Costa,Andrew C Heath,Luigi Ferrucci,Dan Rujescu,Gonneke Willemsen,Najaf Amin,Wendy S Slutske,Cornelia M Van Duijn,Jean-Paul Fox,Andres Metspalu,Igor Rudan,Jari Lahti,Antonio Terracciano,Jüri Allik,David C Smyth,Robert F Krueger,William G Iacono,Marina Ciullo,Ian J Deary,Alan J Gow,Markus Jokela,Richard J Rose,Harry Campbell,James F Wilson,Jaime Derringer,Matt Mcgue,Kauko Heikkilä,Narelle K Hansell,Marleen H M De Moor,Angelina R Sutin,Jasper Wouda,Terho Lehtimäki,Josine Verhagen,Alejandro Arias Vasquez,Anu Realo,Kate Northstone,Caroline Hayward,Teresa Nutile,Michelle Luciano,Ulla Broms,Lea Pulkkinen,John M Starr,Patrik K E Magnusson,Nancy L Pedersen,Klaasjan G Ouwens,David M Evans,Erik Pettersson,Jaakko Kaprio,Daniela Ruggiero,Ilkka Seppälä,Johan G Eriksson,Bettina Konte,Aarno Palotie,Arpana Agrawal,Gerard Van Grootheest,Tõnu Esko,Eero Vuoksimaa,Rossella Sorice,Toshiko Tanaka,Pamela A F Madden,Annette M Hartmann,Nicholas G Martin,Karin J H Verweij,Katri Räikkönen,Brenda W J H Penninx,Lina Zgaga,Eco J C De Geus,Alexander Viktorin,George Davey Smith,Olli T Raitakari,Dorret I Boomsma,Margaret J Wright,Laura Pulkki-Råbäck ,Anu‐Katriina Pesonen ,Lambertus A Kiemeney ,Sita H Vermeulen ,Elisabeth Widén ,Liisa Keltikangas‐Järvinen ,Lindsay Matteson ,Barbara Franke ,Richard A Grucza ,J.g.e Janzing ,Laura J Bierut ,Ozren Polašek ,Jennifer E Huffman

doi:10.1007/s10519-014-9654-x

Abstract

Mega- or meta-analytic studies (e.g. genome-wide association studies) are increasingly used in behavior genetics. An issue in such studies is that phenotypes are often measured by different instruments across study cohorts, requiring harmonization of measures so that more powerful fixed effect meta-analyses can be employed. Within the Genetics of Personality Consortium, we demonstrate for two clinically relevant personality traits, Neuroticism and Extraversion, how Item-Response Theory (IRT) can be applied to map item data from different inventories to the same underlying constructs. Personality item data were analyzed in >160,000 individuals from 23 cohorts across Europe, USA and Australia in which Neuroticism and Extraversion were assessed by nine different personality inventories. Results showed that harmonization was very successful for most personality inventories and moderately successful for some. Neuroticism and Extraversion inventories were largely measurement invariant across cohorts, in particular when comparing cohorts from countries where the same language is spoken. The IRT-based scores for Neuroticism and Extraversion were heritable (48 and 49 %, respectively, based on a meta-analysis of six twin cohorts, total N = 29,496 and 29,501 twin pairs, respectively) with a significant part of the heritability due to non-additive genetic factors. For Extraversion, these genetic factors qualitatively differ across sexes. We showed that our IRT method can lead to a large increase in sample size and therefore statistical power. The IRT approach may be applied to any mega- or meta-analytic study in which item-based behavioral measures need to be harmonized.Electronic supplementary materialThe online version of this article (doi:10.1007/s10519-014-9654-x) contains supplementary material, which is available to authorized users.

Highlights

Mega- or meta-analytic studies (e.g. genome-wide association (GWA) studies) are increasingly used in behavior genetics
To assess whether test linking was successful within the seven cohorts that assessed more than one personality inventory, latent scores were computed based on different calibrations
This study examined for Neuroticism and Extraversion personality traits whether measures from different inventories could be harmonized using Item-Response Theory (IRT) test linking

Summary

Introduction

Mega- or meta-analytic studies (e.g. genome-wide association (GWA) studies) are increasingly used in behavior genetics. Because phenotypes have not always been assessed across cohorts (and sometimes not even within cohorts), measures need to be harmonized, that is, phenotypic scores need to be made comparable such that data from individuals who were assessed by different inventories can be compared meaningfully. Such harmonization enables fixed effect meta-analytic analyses (Hedges and Vevea 1998). GWA studies for psychiatric disorders have led to important discoveries, but for many disorders, individual variants typically explain less than 1 % of the heritability, in unison they can explain quite a large proportion of phenotypic variation (Craddock et al 2008; Lee et al 2013; Ripke et al 2013; Sullivan et al 2012).

Methods

Results

Conclusion