A peer-reviewed, evidence-based journal for clinicians in the field of neuroscience
Issue link: http://innovationscns.epubxp.com/i/425963
[ V O L U M E 1 1 , N U M B E R 9 – 1 0 , S E P T E M B E R – O C T O B E R 2 0 1 4 ] Innovations in CLINICAL NEUROSCIENCE 97 comparator eight-week trial examining BMS 562086, escitalopram, and placebo for general anxiety disorder (GAD). Eighty-two subjects c ompleted 297 S-STS ratings across study time points. Of these, 61 had a baseline and at least one post-baseline follow-up S-STS. Despite the small sample and low occurrence of suicidal ideation in this study, the authors found that the sensitivity of the S-STS (its ability to detect true cases of suicidal ideation or behavior) was very high compared to the rater- administered Hamilton Depression Rating Scale (HAM-D) item #3 (suicide) and in relation to reported suicidal adverse events (AEs) and suicidal serious adverse events (SAEs). The S-STS showed 100- percent sensitivity for identifying subjects with suicidal ideation and behaviors, compared to 63 percent for item 3 (suicide) on the HAM-D. These sensitivity calculations were based on any evidence of suicidal ideation or behaviors from review of AEs, SAEs, HAM-D item #3, or a S-STS score greater than 0. Preti et al 15 evaluated the reliability (internal consistency and test-retest stability) as well as the convergent validity and divergent validity of the self-rated S-STS using a cross- sectional study in a nonclinical sample of 303 college students aged 18 to 40 years (mean age of 23.5 years). Within this sample, 29.4 percent had experienced passive suicidal ideation, 20.4 percent had experienced active suicidal ideation, 4.3 percent had planned a suicide attempt, 1.9 percent had engaged in preparatory suicidal behavior, 5.9 percent had engaged in self-injurious behavior without intent to kill themselves, and 2.6 percent had made a suicide attempt. Sixty subjects were contacted for retesting 4 to 6 weeks later, and 58 complied. Overall, the authors found "promising evidence on the convergent, divergent, internal consistency, and test-retest stability of the S-STS." Youngstrom et al 16 presented an analysis of the concordance of the S- STS with the C–SSRS in a clinical sample of 196 suicidal subjects drawn from inpatient and emergency room settings at Penn State Milton S. Hershey Medical Center in Hershey, Pennsylvania, USA. The analysis m apped each scale to the C-CASA categories that were adopted in the 2010 FDA Draft Guidance 1 for suicidal assessment. The authors found acceptable concordance between the S-STS and the C–SSRS using Kappa scores. A further analysis using the updated 2012 FDA classification categories suicide ideation and behavior categories is underway by this group from this dataset. The University of Alabama at Birmingham (UAB) conducted a validation study comparing the S-STS and the ISST-Plus 17 with the C-SSRS (treated as the gold standard) in 40 adult subjects identified as having suicidal ideation or behavior with varying degrees of severity recruited from inpatient, outpatient, and emergency room settings. The data were analyzed using mapping to both the 2010 and the 2012 FDA Draft Guidances 1 ,2 suicidality categories. The results were reported at five national scientific meetings in poster and oral presentations 11,18 and have generated one published article so far. 11 Subjects were rated on all three scales on the same day and were assigned to each scale in a random sequence order. The S-STS clinician-rated, S-STS patient- rated, and S-STS reconciliation versions were used in this study and were part of this validation analysis. In essence, the S-STS and ISST-Plus were very similar to each other in the agreement scores and diverged very little from each other, although the two scales appear different on face inspection and method of use and are derived from very different sources (Figure 3). However both the ISST- Plus and S-STS (in all 3 versions) showed marked differences with the C-SSRS on most of the suicidal ideation items, which are the essential core of most suicide assessments. The reasons for these differences are the subject of of another paper. 11 The problem here is the flawed navigation instruction for suicidal ideation on the C–SSRS relating to item 2, the flawed application of a Guttman Scaling procedure on the C–SSRS, and the related type I and type II errors on the C–SSRS that lead t o under-endorsement of some suicidal phenomena and over- endorsement of others. 19 For example, the true rates of non-specific active suicidal ideation are inflated on the C- SSRS compared to reality. The Guttman Scaling assumption as used in the C–SSRS is not an optimal, comprehensive, reproducible, or generalizable assumption to model suicidal ideation and behavior. See Appendix H for further details on why Guttman scaling is not optimal for use in the design of a suicidality scale. C- SSRS datasets cannot be merged with ISST-Plus and S-STS datasets because of these design flaws. Evidence of reliability. To date, there are no published data supporting inter-rater reliability of the S-STS; however, the UAB clinical research team conducted a test-retest reliability and inter-rater reliability study in conjunction with the previously described validation study, 11 which confirmed the S-STS reliability. This study is currently being prepared into a manuscript for publication. 18 COMPARISON OF CLINICIAN- RATED, SELF-REPORT, AND RECONCILED VERSIONS OF THE S-STS The correspondence between the clinician-rated, self-report (patient- rated), and reconciled versions of the S-STS appears to be high. As shown in Figure 4, the three versions showed similar patterns in the way they mapped or failed to map to the C–SSRS and the FDA-CASA 2012 categories in the UAB study described above. 11 The reconciliation version was more similar to the clinician-rated version than to the self-rated version, but the interpretation of this finding is complex. While it is tempting to assume that the clinician-rated and the reconciliation versions best reflect what actually occurred, one should not assume this is always or even usually the case.