The Editor’s choice June 2020

written by Jess G. Fiedorowicz, Editor-in-Chief, Journal of Psychosomatic Research, June 2020

Jess Fiedorowicz, JPR’s Editor-in-Chief

The Editor’s Choice – Modeling a Type D Personality Effect

I have been excited about this quarters Editor’s Choice from the Journal of Psychosomatic Research since the moment I reviewed the original submission.  This single author paper by Tilburg University PhD student Paul Lodder also earned him recognition as the 2020 European Association of Psychosomatic Medicine Young Investigator Award.  He was selected from a field of 14 competitive applications.  This paper is different than most papers published in our journal, which tend to focus on studies of clinical samples.  Paul used statistical simulation to assess the optimal way to model synergy with application to the Type D personality construct [1].  Statistical simulations involve the creation of large numbers of datasets to compare the performance of specific methods.  Lodder tested the most commonly used methods to measure a Type D effect to estimate their bias, power, and false positive rate.

In his classic paper, published a few months before I was born, Kenneth Rothman described synergy as when “the risk attributable to a combined exposure exceeds the sum of the risks attributable to each exposure separately [2].”  Thus, synergy involves an interaction with more than additive effects.  In the case of Type D personality, a construct that combines the two personality traits negative affectivity (NA) and social inhibition (SI), our interest in the construct ultimately rests on the assumption of synergy.  Should the NA and SI components simply exert additive effects, then there is no compelling reason to conceptualize them as anything more than independent risk factors, just as we might conceptualize most risk factors for any given condition.  Should the NA and SI components exert less than additive effects (i.e., antagonize each other), then the combined construct is devalued.  In his statistical simulation, Paul Lodder showed that commonly used methods to assess a Type D personality effect commonly produce a false positive Type D effect when only one of the main effects explains the association.  These biased categorical methods classify people into either two or four personality groups based on dichotomization of the continuous NA and SI measures.  Mr. Lodder kindly provided a brief summary of his paper with the following:

Type D personality, a combination of negative affectivity (NA) and social inhibition (SI), has been associated with various adverse outcomes. Type D effects can be analyzed according to several methods. In this study, a computer simulation revealed that classifying people in personality groups produced false positive Type D effects when only NA or SI was related to an outcome. This problem did not occur when modeling Type D using the continuous NA and SI scores. Reanalysis of two published studies showed that significant Type D effects based on personality groups were no longer significant when using the continuous method.

Last year, a publication in the journal nicely illustrated this principle.  Lee and colleagues found an association between Type D personality and in-stent neoatherosclerosis in patients with cardiovascular disease.  When modeled as a categorical construct, Type D personality was strongly associated with in-stent neoatherosclerosis (OR 2.99, 95% CI 1.35-6.63).  However, when modeled as the continuous NA and SI scores with their interaction, only NA was associated (OR for z-score of 1.86, 95% CI 1.26-2.74).  The NA*SI interaction was not significant and in the direction of less than additive effects [3].   In another analysis published before I assumed the role of Editor-in-Chief, Schoormans and colleagues found an association between Type D personality and all-cause mortality in colorectal cancer survivors (aHR=1.7, 95% CI 1.3-2.4).  However, when modeling personality components individually as continuous measures, there were no effects of either NA or SI and their interaction was not significant.  Continuous measures of NA were, however, significantly associated in some subgroups (those 70 years of age and older, men, and comorbid cardiovascular disease) [4].

The Lodder paper provides a compelling explanation for the results of these recent publications by Lee et al. and Schoormans et al. in the journal.  His findings also resonated with the results of a series of analyses I’ve embarked on.  Mixed states, involving symptoms of both depression and mania, have long been considered to convey an especially high risk of suicide in bipolar disorder, largely based on cross-sectional studies showing a high frequency of past suicide attempts in those with such states [5].  In an analysis of a prospective cohort of 429 participants with bipolar disorder followed for a mean of 18 years, we found no evidence of synergistic effects between manic and depressive symptoms on suicidal behavior or completions with a non-significant interaction in a less than additive direction.  Those with a history of mixed states indeed engaged in more suicidal behavior, however, this was due to a more persistent course of depressive symptomatology over follow-up [6].  We followed this up with an analysis of real-world clinical data from the U.S. National Network of Depression Centers, modeling depressive and manic symptoms with their interaction using continuous rating scales, this time for the outcome of suicidal ideation or attempts.  The interaction was again not significant and again in a less than additive direction [7].  This latter analysis has been replicated in a larger sample although is not yet published.  In all three analyses, it was the main effect of depressive symptoms that was most strongly associated with suicide-related outcomes.

For those interested in exploring the Lodder simulation, there is an open source online application available:  https://plodder.shinyapps.io/Type_D_effect_simulation/.  Paul Lodder also produces music.  If you chose to play around with his online simulation, consider also having some of his music play in the background.  You can find a sampling at https://www.paultwin.com/.  Returning to the field of psychosomatics, I hope to see investigators tune their future analyses in the key of Lodder’s findings, by modeling the synergy between two dimensional variables as a continuous interaction effect, rather than creating dichotomized groupings based on whether participants score above a particular threshold on both continuous variables.  If you plan to embark on any such analysis, please send it our way, where your rigor will be appreciated.  If you do send in a paper looking at Type D personality as an exposure of interest, we will likely expect at least a secondary analysis looking at the components as continuous measures and modeling their interaction.  Hopefully, Lodder’s work encourages you to make this the primary analysis in your statistical analysis plan.

 

References:

[1] P. Lodder, Modeling synergy: How to assess a Type D personality effect, J Psychosom Res 132 (2020) 109990.

[2] K.J. Rothman, Synergy and antagonism in cause-effect relationships, Am J Epidemiol 99(6) (1974) 385-8.

[3] R. Lee, H. Yu, X. Gao, J. Cao, H. Tao, B. Yu, Y. Wang, P. Lin, The negative affectivity dimension of Type D personality is associated with in-stent neoatherosclerosis in coronary patients with percutaneous coronary intervention: An optical coherence tomography study, J Psychosom Res 120 (2019) 20-28.

[4] D. Schoormans, O. Husson, J. Denollet, F. Mols, Is Type D personality a risk factor for all-cause mortality? A prospective population-based study among 2625 colorectal cancer survivors from the PROFILES registry, J Psychosom Res 96 (2017) 76-83.

[5] A. Schaffer, E.T. Isometsa, L. Tondo, H.M. D, G. Turecki, C. Reis, F. Cassidy, M. Sinyor, J.M. Azorin, L.V. Kessing, K. Ha, T. Goldstein, A. Weizman, A. Beautrais, Y.H. Chou, N. Diazgranados, A.J. Levitt, C.A. Zarate, Jr., Z. Rihmer, L.N. Yatham, International Society for Bipolar Disorders Task Force on Suicide: meta-analyses and meta-regression of correlates of suicide attempts and suicide deaths in bipolar disorder, Bipolar Disord 17(1) (2015) 1-16.

[6] J.E. Persons, W.H. Coryell, D.A. Solomon, M.B. Keller, J. Endicott, J.G. Fiedorowicz, Mixed state and suicide: Is the effect of mixed state on suicidal behavior more than the sum of its parts?, Bipolar Disord 20(1) (2018) 35-41.

[7] J.G. Fiedorowicz, J.E. Persons, S. Assari, M.J. Ostacher, P. Zandi, P.W. Wang, M.E. Thase, M.A. Frye, W. Coryell, G. of the National Network of Depression Centers Bipolar Disorders Interest, Depressive symptoms carry an increased risk for suicidal ideation and behavior in bipolar disorder without any additional contribution of mixed symptoms, J Affect Disord 246 (2019) 775-782.

 

Find out more:  https://www.journals.elsevier.com/journal-of-psychosomatic-research/editors-choice


The following article has been selected by our Editor-in-Chief:

Modeling synergy: How to assess a Type D personality effect

By Paul Lodder

Vol. 132, May 2020

Corresponding Author’s Commentary

Type D personality, a combination of negative affectivity (NA) and social inhibition (SI), has been associated with various adverse outcomes. Type D effects can be analyzed according to several methods. In this study, a computer simulation revealed that classifying people in personality groups produced false positive Type D effects when only NA or SI was related to an outcome. This problem did not occur when modeling Type D using the continuous NA and SI scores. Reanalysis of two published studies showed that significant Type D effects based on personality groups were no longer significant when using the continuous method.

The above is currently in the process of being added to our journal homepage at https://www.journals.elsevier.com/journal-of-psychosomatic-research/editors-choice


0 Comments

Leave a Reply