Conducting Delirium Research

Transcription

1 Optimizing Clinical Trials When Conducting Research Research funding: Disclosure NHLBI, NIA, AstraZeneca John W. Devlin, PharmD, FCCP, FCCM, Professor of Pharmacy, Northeastern University Scientific Staff, Division of Pulmonary and Critical Care Medicine, Staff of and Care Tufts Medical Center, Boston, Massachusetts 1 2 Outline Study Protocol Considerations: Objective Design Sample Confounding variables Intervention ti Outcome(s) Sample size Data management and analysis DSMB 3 Pain Fear Depression Mobilization and Rehab Reduced d Functionality Mortality Chronic Pain Agitation Altered Sleep ICU memories Return to Independence Persistent Cognitive Defects Increased healthcare costs Family stress 4 Background is a common sequelae of critical illness and worsens both ICU and post-icu outcomes Ouimet S. et al. Intens Care Med 2007; 33: Ely EW et al. JAMA 2001; 78:221 Pandharipande P, et al. N Engl J Med 2014; 370: Intensive Care Screening Checklist (ICDSC) 1. Altered level l of consciousness 2. Inattention = 4 of 8 domains 3. Disorientation - inattention must be present 4. Hallucinations Subsyndromal delirium = 1-3 of 8 domains 5. Psychomotor agitation or retardation Neither nor SSD = 0 of 8 domains 6. Inappropriate speech 7. Sleep/wake cycle disturbances 8. Symptom fluctuation ti An ICU patient who develops SSD (vs. a patient who develops neither SSD nor delirium) is more than 4x as likely to die in the ICU, spend more time in both the ICU and the hospital and be transferred to a SNF (vs. home) Bergeron N. et al. Intens Care Med 2001; 27: Ouimet S et al. Intens Care Med 2007; 33: Role of Antipsychotics for Prevention in the ICU? Strong evidence supports the routine use of non-pharmacologic delirium prevention strategies in the ICU Will a pharmacologic delirium prevention strategy provide additional benefit? Schweickert WD, et al. Lancet. 2009;373: Girard TD et al. Lancet 2008; 371: Kamdar B et al. Crit Care Med 2013; 41: Peri-operative antipsychotic administration may reduce delirium burden in non-critically ill populations Prakanrattana U Anaesth Intens Care 2007; 35: Kaneko T, et al., Yonago Acta Med 1999; 42: Kalisvaart KJ et al. J Am Geriatr Soc. 2005;53: Larsen KA et al. Psychosomatics 2010;51: Hakim SH, et al. Anesthesiology 2012; 116:975-6 Wang W et al. Crit Care Med 2012; 40: One uncontrolled study suggests that t haloperidol l use over the course of the may reduce delirium and mortality van Boorgaard M, et al. Crit Care 2013; 17:R9 However, two RCTs, suggest that use of haloperidol in critically ill patients (with delirium or at high risk for delirium) does not influence patient outcome. MENDS Girard TD et al. Crit Care Med 2010; 38: HOPE-ICU Page VJ et al Lancet Respir Dis Aug

2 Hypothesis Administration of scheduled, lowof scheduled low dose, IV haloperidol in mechanically ventilated, critically ill adults with subsyndromalill adults subsyndromal delirium will reduce the conversion to delirium. NIA 1R15AG A1 #NCT Al-Qadheeb NS et al. Crit Care Med 2015 (ahead of press Nov 4) 8 Defining the Study Objective(s) Should include: an expression describing the overall approach to assess, to compare, to determine aclear description of the intervention Note: if medication should include dose, route and frequency include dose and frequency the disease being evaluated the patient population being evaluated the general purpose of study efficacy, safety, quality of life The primary study outcome Limit the number of secondary objectives Too many multiple comparisons will affect statistical rigor of study If pilot study, additional objectives may include feasibility, evaluation of multiple y y p methods to measure primary outcome, determination of variance around the mean (to guide future sample size calculation) 9 Randomized Study Design Only method to estimate causality Best way to account for confounding and bias to account confounding and Cohort Non time dependent Time dependent Use of a Markov model(s) that incorporates multinomial logistic regression analysis Both baseline and daily variables 10 Considerations when transitioning from one mental state to another ICU Day x ICU Day x+1 Awake, not delirious Awake, not delirious Coma Coma Discharge Awake, no delirium Baseline delirium Day # 1 Awake, no delirium Day #2 Coma Day #3 Day #4 Awake, no delirium Day #5 risk factors ICU Stay considered no Drug X Drug X Drug X Drug X No Drug X Death infusion started infusion infusion 12

3 What about bias? Immortal time bias When a period of 'immortal time' is excluded Systematic deviation in the variable of interest Death Follow up time Critical illness Immortal time Study Sample Who is your sample? sample? This is your inclusion criteria A broader sample can be an advantage Single center vs. Multi center? center? Pragmatism vs. control of confounding factors exclusion criteria = external validity Exclusion criteriacriteria IRB required (pregnancy, prisoner, consent not available etc) Factors that could increase safety concerns IRB has big influence (if medication intervention i i t IRB uses package insert) Factors that might confound ability to measure clinical response This is most common reason external validity is low in many delirium studies Consider a prior stratification at time of randomization to account for i tifi ti tti f d i ti t tf confounders At the very least consider SELECTIVE post hoc secondary analysis (that is defined a priori) Use your DSMB as tool to influence an IRB that may be excessively risk averse All mechanically ventilated adults admitted to 3 different ICUs (medical n=2; surgical n=1) were evaluated q12h with SAS and the ICDSC for up to 3 days from the time of ICU admission: - Important way to reduce inclusion bias - Potential treatment effect of haloperidol to prevent delirium felt to decrease over time Among the 481 patients who had SSD only 68 were randomized = 14.1!! The Tufts MC IRB started off making us exclude patients >/=80 years from trial - DSMB helped get this up to 85 years Important to carefully consider and define the criteria for stopping study intervention

4 Even in Smaller Pilot Studies, Randomization Usually Leads to Study Groups that are Similar at Baseline Haloperidol Placebo Age, yrs 61.7 ± ± 14.9 Male, % APACHE II, at study enrollment 19 [17-23] 20 [17-24] Medical ICU days before enrollment 1[02] [0-2] 1[02] [0-2] IQCODE score 3 [3-3] 3 [3-3] Pre-Deliric score, % 51 [36-75] 48 [38-71] Cont. IV sedation at randomization, % Propofol None Continuous IV opioid at randomization, % SAS at study entry 3 [3-3] 3 [3-3] ICDSC score at study entry 2 [1-2] 2 [2-2] Median [Interquartile range] All differences p Even in Smaller Pilot Studies, Randomization Usually Leads to Study Groups that are Similar at Baseline Haloperidol Placebo Age, yrs 61.7 ± ± 14.9 Male, % APACHE II, at study enrollment 19 [17-23] 20 [17-24] If there Medical is an important subgroup of patients who may respond 67.6 to intervention 73.5 differently: ICU days Incorporate before enrollment stratification prior to randomization1[02] [0-2] 1[02] [0-2] IQCODE score 3 [3-3] 3 [3-3] If there Pre-Deliric is major score, potential % confounding factor that could influence 51 [36-75] response 48 [38-71] to the intervention: A priori define a subgroup analysis to evaluate whether the Cont. IV sedation at randomization, % primary outcome differs between the two different patient groups Propofol None Continuous IV opioid at randomization, % SAS at study entry 3 [3-3] 3 [3-3] ICDSC score at study entry 2 [1-2] 2 [2-2] Median [Interquartile range] All differences p Intervention(s) Need to carefully describe and justify Remember: this is the one thing you control! Is a placebo feasible and/or available? Are there attributes/effects of the intervention that could allow treatment allocation to be detected? t t ti t t t d? Bedside clinicians love to try and guess allocation If bedside clinician is expected to administer intervention (e.g., medication) then cannot be time consuming and must be within then cannot time consuming and must scope of practice Make sure the study medication is incorporated in existing drug distribution/administration system. Use the CPOE system to control for the administration of confounding medications Non study antipsychotics; dexmedetomidine 21 Outcomes: as the Primary Outcome? Prevention of Incidence ICU days without it Time to first delirium episode Duration of first delirium episode Severity of the delirium that occurs Motoric subtypes Treatment of Time to first resolution Duration of delirium Days without delirium in the ICU Consider each ICU day as an individual measurement dua e e and report the daily OR from delirium to non delirium **Important to ask: But should a characterization of delirium be my primary outcome? Do think about the primary outcomes most important to the patient (and their families) Neufeld KJ, et al. Am J Geriatr Psychiatry 2014; 22:1513 Yang FM, et al. BMC Med Research 22 Method 2013; 13:8 Davis DHJ, et al. Am J Geriatr Psych 2013 Consider the Rigor of the Assessment 1. Is once daily delirium assessment enough? 2. Will CAM-ICU assessment by bedside clinicians alone miss delirium? 3. Will the presence of sedation influence delirium detection? Rapidly Reversible vs. Persistent UMC-Utrecht:Assessment of delirium every 8 hours < 24 hours maximum RASS (3hrly) -4 or -5 YES Unable to assess < 24 hours positive CAM-ICU bedside nurse YES Delirious < 24 hours start haloperidol/quetiapine YES Delirious RASS -3 CAM-ICU - RASS -3 CAM-ICU + Evaluation nursing charts for anxiety, hallucinations, disorientation AND impaired/fluctuating consciousness AND/OR initiation of scheduled antipsychotic therapy No YES Delirious Delirious Test characteristics compared to delirium expert team: Sensitivity = 0.75 Specificity = 0.88 Interrater agreement= 0.94 Zaal et al. J Crit Care mechanically ventilated MICU adults managed with propofol/fentanyl if CAM-ICU + = rapidly reversible, sedation-related delirium Sedation Interruption if CAM-ICU + = persistent delirium Patient Outcome Patel SB et al. AJRCCM 2014; 189:

5 Rapidly Reversible vs. Persistent - Rapidly reversible delirium i (before sedation interruption) ti is 10.5 times more likely than persistent delirium (after sedation interruption) - This relationship not affected by the specific ICU admission day or the presence of delirium risk factors (e.g., age, severity of illness, corticosteroid use) Mortality (%) How were these issues accounted for in the haldol study? ICSDC used in study ICUs for more than 10 years y All nurses receive formal ICDSC re education every 6 months that includes the correct evaluation of at least two patients (vs. a clinical nurse educator) ti t i l t Well established DA SBT protocol ICDSC assessment protocolized to occur after sedation occur DA (ie. when patient maximally awake) All positive delirium assessments (ICDSC 4) were confirmed with a study investigator (using ICDSC) and a consultant psychiatrist using DSM IV criteria Patel SB et al. AJRCCM 2014; 189:

6 Special Considerations When Evaluating Safety Safety Consider protocolizing the detection and management of likely safety concerns Want to shown the IRB you are serious about safety Do not want to remove a patient from the study if potential safety concern could be caused by non study factor(s) or it may resolve with intervention () Clearly understand criteria to differentiate serious adverse events from non serious adverse events and reporting of each Is an SAE expected or not expected? Unblind treatment allocation in only very rare situations and only inform clinical team of allocation of allocation Study Completion Considerations: What s your elevator speech? Study implementation Data collection Clinical team engagement Informed consent Patient and family engagement Publication 33