1
January 16, 2015 / March 10 2015
Quetiapine versus aripiprazole in children and adolescents with psychosis - protocol for the randomised, blinded clinical Tolerability and Efficacy of Antipsychotics (TEA) trial
Statistical analysis plan.
Trial registration: ClinicalTrials.gov NCT01119014
Sponsor Anne Katrine Pagsberg
Background
The Tolerability and Efficacy of Antipsychotics (TEA) trial aims at expandingthe evidence base for choices between antipsychotics for the treatment of psychosis in children and adolescents. In this randomised clinical trial (RCT),we investigatebenefits and harms of quetiapine versus aripiprazole in early-onset psychosis.The TEA trial is a Danish investigator-initiated, independently funded, blinded, multi-center superiority RCT comparing quetiapine versus aripiprazole for treatment of psychosis in patients aged 12-17. The primary outcome is the Positive and Negative Syndrome Scale (PANSS) positive score after 12 weeks of treatment. Secondary outcomes include several efficacy and safety measures. According to the International Conference on Harmonisation (ICH) of Good Clinical Practice (GCP), it is recommended that clinical trials should be analyzed according to a pre-specified plan (Dwan 2008) in order to prevent outcome reporting bias and data driven analysis results.Here we describe the TEA trial statistical analysis plan that has been finalized while the database was still locked and before data analysis was commenced.
Outcomes
As defined in the TEA study protocol (Pagsberg et al. 2014), the data retrieved in the TEA RCT are defined as primary and secondary outcomes.The research hypothesis for the TEA trial is that there is a difference between quetiapine and aripiprazole after 12 weeks of treatment in early onset psychosis of at least 3.0 on the primary outcome measure Positive and Negative Syndrome Scale (PANSS) positive score, and that aripiprazole causes more akathisia than quetiapine; and quetiapine causes more sedation, weight gain, and metabolic adverse effects than aripiprazole. We therefore test if we can reject the null hypothesis that there are no significant differences between the therapies on PANSS positive score, ratings of akathisia, sedation, weight gain and metabolic adverse effects.Based on these hypotheses, the TEA trial has a single primary and four key secondary outcomes (see table 1, item 1-5). The first published report of the trial results will focus on the five specified hypotheses, but also report additional secondary outcomes (table 1, items 6-27), whichare statistically handled as explorative outcomes.Apart from report on the specific harms mentioned above, adverse events are, in accordance with the UKU (Udvalget for Kliniske Undersøgelser) scale, categorized in four types of adverse events, i.e., psychological, neurological, autonomous, or other of each of the adverse events (AE),adverse reactions (AR),serious adverse events (SAE),serious adverse reactions(SAR), and serious unexpected adverse reactions (SUSAR). Each of the combinations are recorded in each patient as the number of occurrences from week 0 and onwards (counts) and as the value of a binary variable (1 if the combination occurred at least once and 0 otherwise).
While the primary and aforementioned four key and additional secondary outcomes are analysed and reported in the main outcome report, further hypothesis generating outcomes of the TEA trial (i.e. cognition, health-related quality of life, detailed analysis of adverse events patterns and predictive validity of early response) will be addressed in separate publications. This also includes validation of specific outcome measures in a healthy matched control group. Furthermore, we will explore long-term effectsin a 52 weeks follow up after randomisation, during which the participants are treated openly from week 12 to week 52.
Statistical analyses
Analyses
The analyses are intention-to treat with two-sided tests. Five types of statistical analyses will be applied: regression using mixed model with repeated measures (type 1), regression using the generalized linear mixed model with repeated measures (type 2), generalized linear model (type 3),logistic regression (type 4), and non-parametric analysis (Mann-Whitney) (type 5). When type 1 or 2 are used the outcomes will be evaluated as the difference between the mean square estimated means in the two intervention groups at 12 weeks following start of intervention.
The type 1 through 4 analyses will be adjusted by the protocol specified stratification variables ((1) age > 14 years, yes/no; (2) PANSS positive score > 20 yes/no), and in addition (3) by centre (the centres from Zealand, Fyn and Jutland will be combined into one centre and the centres from The Capital Region into another, due to few patients in the regions outside the Capital Region) and (4) by the baseline value of the outcome (when relevant and available). In the analysis of longitudinal data, the mixed model with repeated measures (type 1) will be used for continuous quantities. Forbinary outcomes, the generalized linear mixed model (SAS) (type 2) will be used. A model with random intercept and binomial distribution will be used and method = QUAD to get true maximum likelihood estimates rather than pseudo ML and covtest statement to test the null hypothesis that the random effects variance is equal to 0.The models will include as covariates the intervention indicator (I), the adjusting variables (see above), time, and the interaction between time and I. If inspection of the mean value trajectory over time in the two intervention groups suggests a non-linear time course the model will be appropriately expanded, e.g., with the covariate time2. For continuous outcomes the best fitting of the covariance matrices: unstructured, compound symmetric or power function will be used. The Bayesian information criterion and visual inspection of the correlation matrices will be used for this assessment. If a model fit is deemed less than satisfactory, the analysis will be supplemented by a non-parametric test (Mann-Whitney) of the difference between the two distributions at 12 weeks and the results discussed.
As an exploratory analysis of the primary outcome, the interaction between intervention indicator and each of the stratification variables and the centre variable will be conducted. If P <0.05 corresponding exploratory subgroup analyses will be conducted.
Missing values of primary and secondary outcomes
If only outcome values are missing and there are no good auxiliary variables (|r| > 0.40),a complete case analysis will be done (Carpenter JR and Kenward MG. 2013).The covariates will by definition not be missing for the primary outcome since its baseline value is used to define a stratification variable. The missingness of the baseline value will be assessed individually for each of the five specific secondary outcomes. Multiple imputation of the baseline value will be conducted using SAS 9.3.
Sensitivity analysis of the primary outcome
A modified, previously published method (Risom et al 2013) for a worst-case scenario analysis of an outcome measured thrice following start of intervention will be applied. The original method involves single imputation of missing values. In the modified version, the resulting uncertainty will be taken into consideration by using the larger of the two standard errors obtained in the primary analysis and in the sensitivity analysis when assessing the difference between the mean values estimated at 12 weeks. The raw P value of the sensitivity analysis will be reported.
Due to our inclusion of patients within a broad diagnostic spectrum of psychotic disorder and with illness severity ranging from mildly to severely ill, we will perform subgroup analyses for patients with schizophrenia-spectrum psychosis and for patients with at least moderate illness severity (PANSS total above or equal to 80). This will allow us to compare results with other samples of early onset schizophrenia and validate the drug effect in patients for which they were originally designed. Furthermore, it will allow us to address the fact that there is clear indication that lower baseline scores make patients more likely to reach remission that is based on a maximum allowed baseline score on certain items and that higher baseline levels are associated with greater possibility for change from baseline and for "responder status" as defined by a certain mean percentage reduction cut off from baseline (Furukawa 2015).
Significance of findings
The primary outcome will be analyzed and interpreted according to a significance level of P<0.05. With respect to the four specific secondary outcomes, P-values below 0.01 will be interpreted as significant findings, whereas P-values above 0.05 are insignificant. P-values in the interval 0.01 to 0.05 will be further discussed, including evaluations of the power provided with the given sample size for the primary outcome.
Literature:
Carpenter J, Kenward M:Multiple Imputation and its Application. Wiley 2013. Page 24-28. ISBN: 978-0-470-74052-1
Dwan K, Altman DG, Arnaiz JA, Bloom J, Chan AW, Cronin E, Decullier E, Easterbrook PJ, Von Elm E, Gamble C, Ghersi D, Ioannidis JP, Simes J, Williamson PR. Systematic review of the empirical evidence of study publication bias and outcome reporting bias. PLoS One. 2008 Aug 28;3(8):e3081. doi: 0.1371/journal.pone.0003081. Review. PMID: 18769481
Furukawa TA, Levine SZ, Tanaka S, Goldberg Y, Samara M, Davis JM, Cipriani A, Leucht S. Initial Severity of Schizophrenia and Efficacy of Antipsychotics: Participant-Level Meta-analysis of 6 Placebo-Controlled Studies. JAMA Psychiatry. 2015 Jan 1;72(1):14-21. doi: 10.1001/jamapsychiatry.2014.2127. PubMed PMID: 25372935
Pagsberg AK, Jeppesen P, Klauber DG, Jensen KG, Rudå D, Stentebjerg-Olesen M, Jantzen P, Rasmussen S, Saldeen EA, Lauritsen MB, Bilenberg N, Stenstrøm AD, Pedersen J, Nyvang L, Madsen S, Lauritsen MB, Vernal DL, Thomsen PH, Paludan J, Werge TM, Winge K, Juul K, Gluud C, Skoog M, Wetterslev J, Jepsen JR, Correll CU, Fink-Jensen A, Fagerlund B.Quetiapine versus aripiprazole in children and adolescents with psychosis--protocol for the randomised, blinded clinical Tolerability and Efficacy of Antipsychotics (TEA) trial.BMC Psychiatry. 2014 Jul 11;14:199. doi: 10.1186/1471-244X-14-199.PMID: 25015535
Risom SS, Zwisler AD, Rasmussen TB, Sibilitz KL, Svendsen JH, Gluud C, Hansen JL, Winkel P, Thygesen LC, Perhonen M, Hansen J, Dunbar SB, Berg SK. The effect of integrated cardiac rehabilitation versus treatment as usual for atrial fibrillation patients treated with ablation: the randomised CopenHeartRFA trial protocol.BMJ Open. 2013 Feb 20;3(2). pii: e002377. doi: 10.1136/bmjopen-2012-002377. Print 2013.PMID: 3430599