You are on page 1of 13

Strategies to Reduce Attrition: RCTs in Psychopharmacology

Andrew C. Leon, Ph.D. Weill Cornell Medical College


Funded, in part, by NIH MH060447

Outline

Attrition rates in psychopharmacology Strategies to reduce attrition


Assessment procedures Intention to Treat analyses in psychopharmacology Definition of outcome Intent to Attend

Attrition Interferes with RCT Goals

Smaller N reduces statistical power Limits feasibility and generalizability Magnitude of attrition bias is function of:
Association of attrition with unobserved outcome Attrition rate

Attrition Rates in Psychopharmacology

Antidepressant RCTs submitted to FDA (Khan, 2000, AGP)

PLA 4 week 5 week 6 week 8 week


45 RCTs

Investigational 25% 38% 38% 36%

Active 43% -36% 38%

37% 38% 41% 36%

N > 19,000 subjects ;

mean: 37%

Attrition Rates in Psychopharmacology

FDA Review of Pediatric Antidepressants and Suicidality


(Hammad, 2004)

PLA MDD Anxiety OCD


24 RCTs

Investigational 24% 35% 30%

Active 44%

28% 37% 29%

N > 4400 subjects;

mean: 32%

Attrition Rates in Psychopharmacology

Geriatric RCTs of Antidepressants (Heo, 2007)

# trials

N 8360

PLA 21.7%

Active 28.1%

Geriatric Depression

68

Attrition Rates in Psychopharmacology

Anxiolytic RCTs submitted to FDA (Khan, 2007, Neuropsychopharm)

# trials GAD Panic OCD 8 10 11

N 1037 1433 1936

PLA 30.2% 25.1% 19.3%

Active 31.1% 25.8% 23.4%

Attrition Rates in Psychopharmacology

Reviews of RCTs for Bipolar Disorder


# RCTs Attrition
(Median)

Acute Depression
(7-8 wks)

Kemp et al. 2008 LA Smith et al. BD, 2007 Gao et al. 2009

14 14 5

36% 68% 76%

Maintenance
(12-30 months)

Maintenance
(12-24 months)

Attrition Rates in Psychopharmacology

Antipsychotic RCTs submitted to FDA (Khan, 2007, Neuropsychopharm)


# trials Schizophrenia

N 3483

PLA 59.0%

Active 48.3%

16

Design RCT to Reduce Loss of Subjects


Reduce subject burden

Restrict duration of assessments (quality vs quantity) 2, 3, 4+ hour baseline assessments are not unusual Only include assessments linked to hypotheses

More accessible assessments


Telephone calls, IVR, Home visits

Ethical guidelines protect subjects

Guarantee each subjects right to exit

Design RCT to Reduce Incomplete Data

Differentiate between med termination & study termination Attempt assessments for entire course of RCT - regardless of adherence to study meds Adhere to Intention to Treat: Analyze as randomized

Implemented in psychiatry; strongly resisted by investigators

Truncated assessment confounds attrition and efficacy


Lavori, Neuropsychopharmacology, 1992

Operationalize Outcomes to Embrace Attrition


Standard Outcomes in Psychopharmacology

Response status based on weekly/biweekly severity ratings


HAMD, YMRS, PANSS, PDSS

Alternative Outcomes

CATIE (schizophrenia; Lieberman, NEJM, 2005)


Time until discontinuation of treatment for any cause

Bipolar Maintenance RCTs (Bowden, AGP, 2000 & 2003):


Time to: relapse, meds for symptom worsening, or dropout

LiTMUS (bipolar disorder):


Necessary clinical adjustments for symptom worsening or side effects (#/month)

Alternative Outcomes that Embrace Attrition


Alternative CNS Outcomes SANAD (Epilepsy; Marson, Lancet, 2007):
Time to treatment failure -- stopping med due to inadequate seizure control, intolerable side-effects, or addition of other AED

DATATOP (The Parkinson Study Group, NEJM, 1993):


Time until levodopa to treat emerging disability

Alzheimer's Disease Cooperative Study (Sano, NEJM, 1997):


Time to death; institutionalization; loss of ability to perform 2 (of 3) basic ADLs, severe dementia.

Strategies to Enhance Retention

Psychoeducation for families increased retention in RCT for bipolar disorder (Sherrill, Psychiatric Services, 1997) Incentives for Participants

$$ Reimbursement, newsletters, certificates, postcards

Accommodate Participants needs

Convenient time and place for assessment

Strategies to Enhance Retention

Engage, thank, and reward participants

Strategies to Enhance Retention

Strategies to Enhance Retention

Lithium Treatment - Moderate dose Use Study for Bipolar Disorder: LiTMUS
Ongoing RCT with 6 month course of tx: 12% attrition

Comparator condition Randomized to optimized tx +/- lithium augmentation Reimbursed $50/visit: costs of travel, child care, parking, time burden Intent to Attend items administered with follow-up questions

Predict Dropout:
Baseline:

Intent to Attend

How likely is it that you will complete the study?


unsure (5) very likely (10)

unlikely (0)

Weekly:

How likely is it that you will attend next assessment session?


unlikely (0) unsure (5) very likely (10)

Leon, Demirtas, Hedeker. Clinical Trials, 2007

Intent to Attend

Simple assessment and adds minimal burden. Included in ongoing RCTs

schizophrenia, depression, ptsd, bipolar disorder, substance abuse & panic

Developed to provide a covariate that predicts attrition

* Identify those at risk of attrition.


Accommodate Ss needs with blinded follow-up questions.

Blinded Follow-up Question


If response is less than unsure (<5 ) Would you be willing to share with me some of the reasons that might interfere with your attendance?

Intent to Attend :

How likely is it that you will attend next assessment session?


unlikely (0) unsure (5) very likely (10)

Blinded Follow-up Questions (LiTMUS)


Time of day Time commitment for study visits Forgets appointments/needs reminders Transportation issues Lack of comfort with the study staff Dislike having blood drawn Side effects of medication Too symptomatic to attend

Intent to Attend :

Intent to Attend
Incorporate in sensitivity analyses Value of Intent to Attend will depend on it association with Attrition, Outcome, and Group. Strength of association will likely vary across indications This item could change non-ignorable attrition to ignorable

Psychiatry Drug Division of FDA


Required LOCF until about 5 years ago. Mixed-effects models are now acceptable as primary, but LOCF must be used in sensitivity analyses

Do not exclude Ss with some missing data With ignorable dropout, mixed-effects models can be used for valid inference. Assume attrition explained by observed outcome or covariates Intent to Attend could prove to be a useful predictor of attrition.

Summary
Attrition rates are substantial in psychopharmacology Design RCTs to minimize attrition

Reduce burden of assessments Continue to assess regardless of adherence to study meds Operationalize outcome to incorporate dropout Provide incentives Collect data that predict dropout Accommodate participants needs