Kantojärvi et al. investigated risk factors of personality disorder (PD) in 508 inpatient adolescents [Reference Kantojärvi, Hakko, Riipinen and Riala1]. By follow-up of 9.2 years on average, a total of 57 (11.2%) inpatients were diagnosed as PD in adulthood. Adjusted odds ratio (OR) (95% confidence interval (CI)) of females’ with anxiety disorders in adolescence for PD was 4.39 (2.02–9.53). In addition, OR (95%CI) of males’ hospital treatment from child welfare placements for PD was 3.23 (1.21–8.61). I have two concerns about their study.

First, the same study group examined the effect of bullying behavior in adolescents on the subsequent development of PD by a follow-up of psychiatric inpatients [Reference Antila, Arola, Hakko, Riala, Riipinen and Kantojärvi2]. Adjusted OR (95% CI) of females’ victim of bullying for PD was 3.80 (1.40–10.30). The significance was not observed in males. In addition, Strandholm et al. conducted a prospective study of 140 adolescents with depression to examine the effect of defense mechanisms on the subsequent development of PD in adulthood [Reference Strandholm, Kiviruusu, Karlsson, Miettunen and Marttunen3]. Neurotic and image-distorting defense styles, defenses of displacement, isolation and reaction formation were significant factors of subsequent PD. These studies handled a limited number of patients, and stratified analysis by sex results in the unstable estimate. The minimum number of events per independent variable in a logistic regression analysis is 10 [Reference Novikov, Fund and Freedman4,Reference Stoltzfus5], and the wide range of 95% confidence intervals may reflect an insufficient number of PD as events. To confirm the risk factors of PD in adulthood, continuous survey is required to collect an adequate number of PD.

Second, Perry et al. reviewed the defense mechanism in patients with PD by classifying PD into four types [Reference Perry, Presniak and Olson6]. Regardless of the types of PD, there were consistent associations between defense mechanism and PD. I recommend Kantojärvi et al. stratified analysis by the type of PD in adulthood. As the borderline PD is predominant in female patients [Reference Antila, Arola, Hakko, Riala, Riipinen and Kantojärvi2], Kantojärvi et al. should conduct sensitivity analysis on borderline PD.