Predicting onset of early- and late-treatment resistance in first-episode schizophrenia patients using advanced shrinkage statistical methods in a small sample

Abstract

BackgroundUp to 35% of first-episode schizophrenia patients are estimated to meet criteria for treatment resistant schizophrenia (TRS) over the five years after illness onset (Lally et al., 2016).Emerging evidence demonstrates that around 70% of TRS patients did not respond to antipsychotic medications from the start of first treatment (i.e., early treatment resistant (E-TR) schizophrenia); whereas, the remaining 30% of TRS patients, broadly defined as latetreatment resistant (L-TR) schizophrenia, gradually transition to treatment resistance during the 5-year period, having initially responded to antipsychotic medications (Lally et al., 2016).Given TRS is a major cause of disability and is associated with high social and economic costs (Kennedy et al., 2014), having reliable prediction models for estimating an individual risk for E-TR and L-TR would advance our understanding of patients' risk of developing treatment resistance, especially after a period of favourable response to the ongoing treatment with antipsychotic medications. Development of accurate prediction models for estimating individual, rather than average, risk for a disorder outcome during the illness (Steyerberg et al., 2010), based on available patient characteristics and clinical findings, is generally thought to require large datasets (Califf et al., 1997), mainly when the outcome of interest is a rare event. As they are often unavailable in schizophrenia research, stepwise selection techniques, particularly with the traditional p-value of 0.05, are frequently employed, which tend to provide readily interpretable models (Steyerberg et al., 2000). However, these stepwise selection techniques increase the risk of biased regression coefficients and overfitting (Derksen and Keselman, 1992) potentially leading to development of a prediction model with inaccurate predictions in new cases (Osborne et al., 2012;Steyerberg et al., 1999).Computer intensive statistical learning methods, particularly regularised regression methods (RRMs), have been suggested as optimal methods for clinical and personalised risk prediction (Steyerberg, 2019), especially for small datasets (Steyerberg et al., 2000). Through an

Interplay between polygenic propensity for ageing-related traits and the consumption of fruits and vegetables on future dementia diagnosis

“…The reason for focusing on the previous day is that memory is stronger for recent experiences. Finally, the proportion of missing data in the present study was comparable to other populationrepresentative longitudinal cohort of similar design [21,70,71] and within the range for missForest to handle it efficiently [72,73].…”

Study protocol for the development and internal validation of Schizophrenia Prediction of Resistance to Treatment (SPIRIT): a clinical tool for predicting risk of treatment resistance to antipsychotics in first-episode schizophrenia

“…Therefore, in the two data sources for our model development, we define the TRS cases as those who did not respond to two consecutive antipsychotic medication of adequate dose and for an adequate duration and/or the documented reason for switching antipsychotic medication was due to a lack of therapeutic response 6 7 47–49. An adequate daily dose of antipsychotic medication was defined according to a daily dose of ≥400 mg chlorpromazine equivalence 50.…”

“…A gap of no longer than 14 days was allowed between consecutive treatments. We only included as TRS cases those patients who failed to respond and not those who were intolerant of antipsychotic medications or those who self-discontinued antipsychotic medication 6 7 47–49. Non-response here was measured as not experiencing improvement in symptoms and social functioning in duration of 6 months during the follow-up period 38…”

Study protocol for the development and internal validation of SPIRIT (Schizophrenia Prediction of Resistance to Treatment): A clinical tool for predicting risk of treatment resistance to anti-psychotics in First Episode Schizophrenia

“…[13][14] Therefore, in the two data sources for our model development, we define the TRS cases as those who did not respond to 2 consecutive antipsychotic medication of adequate dose and for an adequate duration and/or the documented reason for switching antipsychotic medication was due to a lack of therapeutic response. [6][7][41][42][43] An adequate daily dose of antipsychotic medication was defined according to a daily dose of ≥400mg chlorpromazine equivalence. [44] We only included as TRS cases those patients who failed to respond and not those who were intolerant of antipsychotic medications or those who self-discontinued antipsychotic medication.…”

“…[44] We only included as TRS cases those patients who failed to respond and not those who were intolerant of antipsychotic medications or those who self-discontinued antipsychotic medication. [6][7][41][42][43] Non-response here was measured as not experiencing improvement in symptoms and social functioning in duration of 6 months during the follow-up period. [45] Study sample…”