Database Studies and the Pitfall of Imputing Complex Clinical Causality

See Reply by Wang et al

This work may not be copied, distributed, displayed, published, reproduced, transmitted, modified, posted, sold, licensed, or used for commercial purposes. By downloading this file, you are agreeing to the publisher’s Terms & Conditions.

Database Studies and the Pitfall of Imputing Complex Clinical Causality

To the Editor: I read with interest the article by Wang et al1 on the impact of drug adherence on oppositional defiant disorder (ODD) and conduct disorder (CD) among Taiwanese children diagnosed with attention-deficit/hyperactivity disorder (ADHD). In it, the authors state that “good drug adherence consistently exerted protective effects on ODD or CD.” If this is true, it is an extremely important finding, and one that should lead treating physicians to urge parents to ensure their children’s ADHD medication compliance by all possible means.

In this database study, the authors found that of 33,835 children diagnosed with ADHD, the one-third of the cohort defined as having good compliance over 90 days of treatment with either methylphenidate or atomoxetine had a lower risk of being diagnosed with ODD or CD over the subsequent 2 years.

The authors’ inference of a protective effect of medication on the development of ODD or CD is flawed for 2 main reasons. First, the authors are aware that the diagnosis rates of both ODD and CD imply substantial underdiagnosis. Even after adding those already diagnosed or diagnosed during the 90-day treatment period to the 4.1% and 4.0% diagnosed with ODD and CD, respectively, over the 2-year follow-up period, the resultant rates of 8.85% and 7.29% are still well below those rates established by interview among Taiwanese children with ADHD, cited as 69% and 33%, respectively.1 The authors acknowledge that it is unclear what selection biases might have affected the rate of ODD and CD diagnoses being documented in their cohort and have listed several possibilities, including frequency of outpatient visits, socioeconomic status, and family dysfunction. Any of these might also have an association with medication compliance.

The second and even more important flaw is that the assertion of a protective effect of medication for subsequent development of both ODD and CD is actually contradicted by the finding that a longer delay in starting medication after a diagnosis of ADHD was also protective. Logically, if consistent medication (good compliance) is protective for ODD and CD, lack of medication (delayed treatment) cannot also be protective.

A far more likely explanation of the correlation between poor compliance with prescribed medication and ODD and CD is that these conditions both predispose to poor compliance. In other words, causality goes in the direction opposite to that suggested by the authors.

A final warning about the pitfall of database studies: even though the study numbers may be appealingly large, extreme caution should be used in imputing complex clinical causality to limited database data of unknown accuracy.

Reference

1. Wang LJ, Lee SY, Chou MC, et al. Impact of drug adherence on oppositional defiant disorder and conduct disorder among patients with attention-deficit/hyperactivity disorder. J Clin Psychiatry. 2018;79(5):17m11784. PubMed CrossRef

Alison S. Poulton, MD (Cantab)a

[email protected]

aDepartment of Pediatrics, Nepean Hospital, Sydney Medical School Nepean, The University of Sydney, Penrith, Australia

Published online: March 12, 2019.

Potential conflicts of interest: Dr Poulton has received personal fees and non-financial support from Shire, outside the submitted work, and holds shares in GSK.

Funding/support: None.

J Clin Psychiatry 2019;80(2):19lr12732

To cite: Poulton AS. Database studies and the pitfall of imputing complex clinical causality. J Clin Psychiatry. 2019;80(2):19lr12732.

To share: https://doi.org/10.4088/JCP.19lr12732

© Copyright 2019 Physicians Postgraduate Press, Inc.