In this post hoc analysis of data from a randomized, double-blind, clinical trial of olanzapine LAI for treatment of acute exacerbation of schizophrenia, early response (Week 4) was found to predict subsequent response (Week 8) with high degrees of sensitivity, specificity, PPV, and NPV, and an overall accuracy level of 79%. The group of patients who met criteria for early response had a trajectory of symptom recovery that was distinct from that of the group who did not meet response criteria. Compared to Early Non-responders, Early Responders had a higher level of symptom severity at baseline as measured by PANSS Total score, and had a greater degree of improvement over baseline at every subsequent time point. By Week 8, improvement by Early Responders was more than double that of Early Non-responders (-37 points vs. -15 points).
The total accuracy of predicting later response to olanzapine LAI based on early response exceeded that seen in studies of oral antipsychotics, including analyses of patients with chronic schizophrenia,[7
]. of patients with first-episode psychosis,[9
]. of a prospective study of patients with chronic schizophrenia,[7
]. and of 5 pooled studies of patients with chronic schizophrenia [6
]. (79% for olanzapine LAI versus 72%, 70%, and 74%, respectively). Much of the increase in total accuracy was due to stronger performance for sensitivity (85% for olanzapine LAI versus 70%, 77%, and 60%, respectively) and PPV (78% for olanzapine LAI versus 63%, 42%, and 54%, respectively). Stronger predictive characteristics in this analysis may have been due to use of a later time point to assess early response; Week 4 for this analysis versus Week 2 in the oral olanzapine studies. Predictions of later response have been shown to improve as the time point for early assessment is delayed,[17
]. and a balance must be struck between improved predictive accuracy and expedient recognition of a need to adjust treatment. In this instance, the later time point was chosen because for many of the patients in this study, 4 weeks represented the interval between injections, and was therefore the earliest point at which treatment could be re-evaluated. Another possible explanation for why the predictive accuracy seen in this study of olanzapine LAI was better than that seen in studies of oral antipsychotics is that with depot formulations, adherence is guaranteed at least through the injection interval. Adherence to oral antipsychotics is not assured. Study populations containing a mix of patients with varying degrees of adherence may have lowered the predictive accuracy of early response/non-response in those studies.
A comparison of rates of study discontinuation between Early Responders and Early Non-responders was included in this analysis because this outcome is associated with increased rates of relapse and psychiatric hospitalization, decreased functional outcome and quality of life, and increased treatment costs [15
]. Moreover, in a response trajectory analysis, treatment response was closely aligned with discontinuation rates [23
]. While the relationship between short-term and longer-term persistence with therapy and the clinical significance of short-term persistence are not known, results presented here suggest that differences in efficacy noted as early as Week 2 may be an important determinant of short-term persistence with treatment.
Early Responders to olanzapine LAI differed significantly from Early Non-responders at baseline by having higher PANSS Total scores. In prior studies of oral antipsychotics for treatment of patients with chronic schizophrenia, Early Responders also had higher baseline PANSS Total scores than Early Non-responders,[6
]. though this difference was not found for patients with first-episode psychosis [9
]. Other baseline discriminators between Early Responders and Early Non-responders have been identified, including differences in PANSS Positive [6
]. and General Psychopathology [6
]. subscores. However, a difference in the PANSS Negative subscore, as was seen in this study, had not been noted previously. It appears that individuals with higher levels of baseline psychopathology may have higher likelihoods of responding to treatment, whether given orally or in a depot formulation. Determining a cut-off for the baseline PANSS score that maximized the predictive value of this assessment tool could help clinicians optimize treatment.
Early improvement in symptoms of schizophrenia in response to olanzapine LAI not only predicted later response in a total measure of symptoms of schizophrenia, but also predicted baseline-to-endpoint improvement across multiple symptom domains and other measures of quality of life, mental health, and social functioning. This study highlights potential benefits to patients, clinicians, families, and payers of early evaluation and identification of patients responding poorly to treatment. These findings echo those of Ascher-Svanum et al.,[4
]. who found that in a post hoc analysis of data from a 1-year, randomized, open-label study of antipsychotic therapy for treatment of schizophrenia, patients who demonstrated response at Week 4 had greater improvement from baseline to Week 8 in measures of mental health, emotional role social functioning, physical functioning, vitality, and a composite mental health score than those who did not respond at Week 4. Likewise, in a randomized, double-blind, flexible-dosed, prospective 12-week study of people with schizophrenia treated with oral risperidone therapy, patients with ≥20% improvement on the PANSS Total score by Week 4 of treatment achieved significantly greater baseline-to-Week 12 improvement across all functional subdomains of the Schizophrenia Objective Functioning Instrument, a validated functional measure including subdomains of living situation, instrumental activity, productive activity, and social functioning [7
Of the patients who did not meet criteria for response at Week 4, 1 in 4 went on to meet the criteria for response at Week 8. A regression model to identify characteristics that distinguish Early Non-responders who do and do not respond at a subsequent time point found that lack of early improvement in 2 PANSS subscale scores - Negative Symptoms and Disorganized Thoughts - predicted continued non-response at endpoint. Specifically, lack of at least a 3-point drop in the Negative subscale score and lack of at least a 3-point drop in the Disorganized Thoughts subscale score at 4 weeks predicted continued non-response. Using this secondary predictive model as a confirming diagnostic test, lack of subsequent response could be predicted with near certainty (NPV = 91%). This secondary prediction model could potentially help in situations where clinicians need reassurance before proceeding with a change in treatment strategy.
This analysis had several limitations that should be noted. First, it focused on a sample that consisted predominantly of inpatients who were acutely, markedly ill at baseline and had a history of multiple relapses. Results cannot be extrapolated to patients with less severe symptomology or who are earlier in their course of illness. This analysis used data from a trial that was 8 weeks in length. Further study is needed to assess this phenomenon over a longer treatment period. Also, the 30-item PANSS used to assess early response/non-response is a lengthy measure that is not used in usual clinical practice, thus potentially limiting the clinical applicability of our findings. However, we have recently identified 6 PANSS items that can be used to reliably assess response/non-response to oral antipsychotics with only a slight loss in predictive accuracy [24
]. Whether an abbreviated assessment such as this can be used to predict longer term response based on early response to olanzapine LAI will need to be determined. Finally, the early and substantial improvement seen in patients with the more severe pathology at baseline may be simple regression toward the mean, but this is a limitation of all analyses of this type.