This study examined the diagnostic value of the French MSI-BPD in a clinical borderline sample and a non-clinical control sample. To our knowledge, this is the first study to assess the MSI-BPD in a population of borderline adolescents. We found that the MSI-BPD demonstrated high diagnostic efficiency (AUC = 0.93), with a cut-off of 5 showing a good correlation with the SIDP IV (sensitivity = 0.87; specificity = 0.85). The sensitivity of our study was among the highest of the available studies and specificity was close to that originally reported by Zanarini and colleagues [5]. Our results are slightly different to those reported in an adolescent and young adult sample by Chanen and colleagues (sensitivity = 0.68, specificity = 0.75, AUC = 0.73), and in an adolescent inpatient population by Noblin and colleagues [12] (sensitivity = 0.71, specificity = 0.65, AUC = 0.73).
We found that a cut-off score of 5 increased the diagnostic efficiency of the MSI-BPD, with particularly improved sensitivity. It should be noted that the cut-off threshold established in the current study is lower than the scores established in the initial validation study [5] (≥7; Zanarini and colleagues) and in several subsequent studies [13, 15, 17, 19]. Our optimal cut-off is below the initially defined threshold of 7 [5], but it's close to a score of 5.5 which was found by Noblin and colleagues [12] in a sample of adolescent inpatients. Several hypotheses may be proposed to explain our results, which are significantly different from other studies using similar samples, i.e. adolescents and young adults [12, 13, 14]. Firstly, the sample sizes of the available studies were smaller overall, ranging from 16 [14] to 31 [12]. Secondly, in the three studies cited above, the control groups were made up of adolescents who were not borderline but who were receiving mental health care whereas our control group was composed of normally developping adolescents. Finally, the severity level of the psychopathology varies considerably because some samples are composed of ambulatory [13,14] and others of hospitalized subjects [12].
The cut-off score recommended to distinguish “cases” from “non-cases” in a questionnaire with a continuous score distribution should depend on the intended use of the scale. If the objective is to identify a relatively homogeneous group of individuals who are very likely to have the disorder being investigated, then a high threshold will be chosen to increase the specificity of the scale and thus reduce the number of false positives. If, however, the objective is a broad screening, the threshold chosen must be lower to increase sensitivity.
The current study has several limitations. Since MSI-BPD is a self-questionnaire, several biases may be present such as the bias of social desirability or social conformity. Subjects with co-morbid Axis I psychiatric disorder were not excluded, nor were subjects who were receiving psychotropic drugs. More specifically, our sample included 37% of adolescents with a major depressive episode. The clinical distinction between depressive disorders and BPD is not easy to make, especially in adolescents, because numerous symptoms overlap [24]. It would have been appropriate to have a group of adolescents with BPD but without depressive disorder, in order to have a clearer idea of the value of the MSI BPD in cases of associated depression. Unfortunately, the size of our sample did not allow us to carry out such analyzes. This limitation should be taken into account for designing future studies.
The cut-off score presented in the current study should be interpreted and used with caution due to these limitations of sampling variability.
In conclusion, this study adds to the growing number of studies suggesting that the MSI-BPD appears to be a feasible screening tool for BPD. Although screening does not replace the use of semi-structured interviews, the MSI-BPD can help to screen borderline subjects and ultimately reduce diagnostic delay.