Innovations In Clinical Neuroscience

NOV-DEC 2017

A peer-reviewed, evidence-based journal for clinicians in the field of neuroscience

Issue link:

Contents of this Issue


Page 50 of 83

51 ICNS INNOVATIONS IN CLINICAL NEUROSCIENCE November-December 2017 • Volume 14 • Number 11–12 O R I G I N A L R E S E A R C H The frequency (37%) with which the Remission criteria are met after treatment might be due to multiple sources: a large placebo response could have driven down some item scores artificially, and these short-term gains could disappear within the six-month Remission timeframe. The symptom changes could be due to score-deflation by raters (i.e., given that the expert panel identified symptoms that should identify overall improvement, it is possible that raters anticipated changes in these symptoms even more so than other symptoms that were not identified by the Remission subgroup, suggesting an affirmation bias). Finally, the Remission subgroup constructed these criteria based on relapse prevention trials, suggesting that these studies might be comparing different trial populations. As shown in Figure 7, the best items to be used for assessment ultimately depend on the symptom severity level, as different items have different utility depending on the illness level of the patient. The items with the most total information for the entire symptom range were those with the largest discrimination parameters: Suspiciousness Persecution, Excitement, Delusions, Tension, Unusual Thought Content, Preoccupation, Hostility, and Conceptual Disorganization. Unusual Thought Content, Conceptual Disorganization, and Delusions were both part of the original Remission criteria. The items providing the most information for patients within the Relief range were calculated using the item characteristic curves. These items were Delusions, Preoccupation, Suspiciousness Persecution, Unusual Thought Content, Conceptual Disorganization, Stereotyped Thinking, Active Social Avoidance, and Lack of Judgment and Insight. Three of these items, Delusions, Unusual Thought Content, and Conceptual Organization, were contained in the original Remission criteria. Lower scores on these items (≤2) were strongly associated with having a low latent trait θ or attaining overall "Symptom Relief." Limitations. There are several limitations to our models. First, we treated active and placebo interventions identically, although, previously, changes had been seen between these interventions with IRT models. 9 Second, patients with reduced symptom severity post- treatment are not necessarily those who are the greatest treatment responders. Third, our models and the Relief criteria proposed here are specific to those enrolled in a drug trial; this might be different than those patients seen clinically and those for whom the original Remission criteria were established. These analyses do not suggest that the Remission criteria do not demarcate those with "an improvement in core signs and symptoms to the extent that any remaining symptoms are of such low intensity that they no longer interfere significantly with behavior and are below the threshold typically utilized in justifying an initial diagnosis of schizophrenia" as originally intended. The core signs were clinically determined for Remission based on their relevance to the disorder and their impact on patients; the IRT analyses is blind to the qualitative impact of symptoms on patients and their life outcomes. Finally, there are also several technical concerns that we need to raise, but do not have space to fully discuss. First, we needed to run the model for a large number of iterations in order to meet the convergence criterion in mirt, possibly the sign of an unstable model. Second, in the bifactor model, the Delusions item tended toward a Heywood case—very high discrimination—that might be affecting other item parameter estimates. Third, the results of the factor analysis (limited information estimation) did not always align perfectly with the result of the IRT analysis (full information estimation). Although we did not expect perfect alignment, interpretation of item quality differed somewhat between solutions, which is potentially another sign of instability. Finally, we note again that some items had cross-loadings in violation of the bifactor model, and the overall fit was modest. CONCLUSION The Remission criteria were drafted based on clinical expertise and grounded in the Diagnostic and Statistical Manual of Mental Disorders (DSM), Fourth Edition. Our results suggest that the subjects attaining "Symptom Relief " measured using the entire PANSS scale are not necessarily Remitted, and those subjects who attain Remission frequently have latent trait scores that suggest moderate-to- severe overall symptom levels. More generally, this disparity between Remission and the latent illness level suggests that the PANSS differentially assesses patients compared to an expert clinician. The term "Symptom Relief " is presented here not just as a description of the symptom range we are profiling, but also to delineate these analyses and these results from Remission. The differences between the Relief and Remission subsets do not suggest that the Relief subset replaces, refutes, or contradicts the original Remission criteria, because the Relief items are selected to answer a different question than the Remission objectives; the Relief criteria provide a method of identifying patients with low general illness severity after the conclusion of a trial, when measured using the entire PANSS. The original Remission criteria were linked to what was considered "active illness" in the DSM, and were considered definitional for schizophrenia and the DSM subtype of "schizophrenia in remission." The findings reported here suggest that A) the PANSS has a strong general factor reflecting overall severity and once this is accounted for there is little additional variation in symptoms and B) satisfying the proposed criteria for Remission is not necessarily associated with low levels of general symptom severity. Together these findings call into question whether rules established to determine satisfaction of the diagnostic criteria (specifically the "A" criteria of the DSM) are truly the best way to characterize whether individuals are still in an "active illness" or have recovered well from an episode illness. While conceptually appealing, there might be value in basing definitions of "remission" or "relief " on our best estimates of general symptom severity. In constructing our criteria for symptom "relief " here, we used a direct measure of the general factor (θ≤1) based on IRT that identifies individuals who, on average, have symptom severity in the "absent" to "minimal" range. This approach more clearly identifies patients with lower severity of symptoms; it remains an empirical question whether application of similar criteria might be more useful clinically, or in clinical trials. Here, the items with the greatest discrimination parameters over the entire illness range overlap considerably, but not perfectly, with the Relief criteria symptoms. This suggests that the utility of any subset of PANSS items ultimately depends on the population being evaluated. It is likely that the

Articles in this issue

Archives of this issue

view archives of Innovations In Clinical Neuroscience - NOV-DEC 2017