PPS Clinical Trials Flashcards
Randomised controlled trial example: IBS
Management: dietary advice, increase 1) soluble fibre and 2) insoluble fibre
Evidence (limited/anectdotal): alleviates or worsens symptoms, so call for objective evidence
Aim: determine effectiveness of increase in dietary fibre: soluble or insoluble
Study design is RANDOMISED PLACEBO CONTROLLED TRIAL
Follow up = 12 weeks from recruitment
What are the outcomes and different types of outcomes we are looking for?
Outcomes:
endpoint- measure effectiveness (benefit)
primary- symptom relief analysed after short and long term treatment for short and long term effectiveness
secondary- IBS symptom severity score
positive- achievement desrirable state (symptom relief, aim to increase)
negative- development disease/condition, IBS symptom severity score, aim to decrease
Randomised controlled trial: Parallel study design
RCT: PSD (between subjects design)
Sample: taken from general practices, select from population so that its representative of population, minimising selection bias
What kind of study is an RCT done over a period of time?
Prospective study
What might give researchers an inflated idea of effectiveness of a treatment in collected data from a RCT?
Percentages increasing/decreasing for certain groups, but not taking into account the numbers of patients dropping out of the study
What is a placebo?
Pharmacologically inert (no active therapeutic ingredients)
Similar appearance and taste to active treatment and administered at the same time and frequency- patients given unmarked sachets
Acts as a control treatment (comparator)
What is the active treatment?
The treatment being investigated- the Intervention.
What happens if intervention not received? (i.e., you’re not in the intervention receiving group- what happens to you)
CONCURRENT CONTROL GROUP to reflect current natural epidemiology (can’t be retrospective because something could have changed in disease’s natural epidemiology)
- Could be standard treatment- active or positive control (eg compare new drug to current standard one already being used, which would be active positive or negative control)
- Or compare active treatment against placebo (negative control)
- No additional treatment- unlikely
What is the main focus of RCT methodology?
Awareness of and minimising biases
What is the purpose of:
Sampling?
Random allocation?
Blinding?
What is essential for inference of causality?
Differences in outcome are due to differences in treatment
this is what you want the study to achieve
eg, primary outcome is adequate relief of pain from condition, any differences between treatment groups should be due to the different treatments and not other reasons- causality. for this need RANDOM ALLOCATION
What does random allocation do?
Eliminates allocation bias -> minimises confounding -> facilitates blinding
What does blinding do?
Double blind- minimises assessor bias and response bias
Especially due to use of placebo as control, unmarked so it looks like active treatment
What is sampling and what is involved?
The population is entire group of individuals of interest- theoretically infinite (statistically) but cost and labour issues in getting a whole population, so we SAMPLE
Sample = subset of population (can be convenient sampling- multicentre)
Is a sample representative of the population?
Generalisation- inferences about population, assumes a lot not totally accurate
There is also SELECTION BIAS -> systemic difference (non random) between sample and population
(eg, are the patients older? greater proportion of females? note that these are the only people that you can make inferences back into the population
What is random allocation and what are the benefits?
Recruit patients and allocate them to a treatment group at random.
Each group is equal so each participant has same probability of 1/3, if three groups like test 1, test 2, control (probability allocation)
BENEFITS:
Eliminates allocation bias
- systemic difference between participants
- can be confused w selection bias
- Not dictated by patients characteristics to avoid bias (subconscious or otherwise)
- minimises confounding (may not eliminate) so treatment groups have similar baseline characteristics and therefore any differences in outcome -> is due to differences in treatment, NOT differences in baseline characteristics
(each group has a similar distribution of ages, gender, ethnicity etc, so it would not mean one group is disproportionately male and so results would result from the different baseline characteristics)
What does random allocation facilitate?
BLINDING
participants do not know what group they are allocated to [neither does researcher (double blind)] -> double blinding minimises bias when reporting and measuring outcomes - this is ASCERTAINMENT BIAS
What is ascertainment bias?
Bias when participants are reporting and researchers measure outcomes
participants- response bias
researchers- assessor bias
How does blinding minimise response bias and assessor bias?
RESPONSE BIAS
- participants bias in reporting outcome
- systematic (not random) difference between patient response and ‘truth’ ie true reports
- may have subjective self report outcomes like symptom relief, to please researchers (if patient knows they have placebo might not be happy and report negatively) underreport/overemphasise effects
ASSESSOR BIAS
- researchers have bias in measuring outcomes
- systematic (not random) difference between actual measurement and ‘truth’
eg, researcher favours a treatment and want to show it as most effective- may elicit response that favours this in interview with patient so inaccurate results, possible dishonesty
blinding hopes to minimise LOSS TO FOLLOW UP, DROP OUT as they don’t know if they’ve been given placebo/treatment
Why do patients drop out of the study?
Feel they have no benefit
Feel worse
Don’t think it’s worth reporting little effects- but this is important because researchers need to know if there is little or no effect to accurately measure population wide effects in the sample
Control treatment
Treatment and placebo responses
TR- observed outcome in intervention groups
components: natural epidemiology, placebo effect, treatment effect
PR- observed outcome in placebo group
components: natural epidemiology, placebo effect
Natural epidemiology
Outcome response without treatment, observed outside the trial
Placebo effect
psychological effect includes:
- response to investigation including therapeutic ritual
- subsequent response to observation and assessment
- response to patient-doctor interaction
Hawthorne effect
part of placebo effect
NON SPECIFIC EFFECT: behavioural change as motivational response to interest, care and attention received from observation and assessment
Treatment effect
Pharmacological effect
QUANTIFY- treatment minus placebo group response
(% minus % placebo = treatment effect %)
What is part of each group response?
Natural epidemiology and placebo effect
Response vs effect
Response is total outcome for group
Treatment response is not the same as treatment EFFECT
same with placebo response and effect
Informed consent
Each participant must undergo informed consent prior to recruitment.
”..consent means a voluntary, uncoerced decision, made by a sufficiently competent or autonomous person on the basis of adequate information and deliberation, to accept rather than reject some proposed course of action.” (Gillon, 1986).
For consent to be valid, information must be understood.
Process involves communication of risk:
- Random allocation;
- Advantages and disadvantages (risk) of treatment.
Ethics of placebo
Declaration of Helsinki
“…..clinical trials that compare new drugs against no treatment (including placebo) when an effective treatment already exists are unethical. Placebos should only be used in cases where there is no proven therapy for the condition under investigation.”
Can give placebo if want to determine how effective treatment is?
“Inclusion of a placebo would not be unethical where for compelling and scientifically sound methodological reasons its use was necessary to determine the efficacy or safety of a prophylactic, diagnostic or therapeutic method; or where a prophylactic, diagnostic or therapeutic method was being investigated for a minor condition and the patients who received placebo would not be subject to any additional risk of serious or irreversible harm.”
Is placebo causing patient any harm?
Cross over trial
Usually for condition alleviated but not cured
Patients get both treatments in random order
Still blinded- but can be open label so patients know which group they’re in but not researchers
Clinical trials summary
Why is outcome assessment important?
Parallel group trial design
One intervention and one control group
Crossover trial design
First phase same as parallel
Then groups swap over for second phase
Factorial design
Parallel groups- multiple intervention groups and a control group
Importance of trial size
Randomized controlled trials are increasingly widely used
• BUT many have been TOO SMALL
• If trial size is small, statistical power limited
– May fail to detect an important intervention effect (`type 2’ error)
– Estimate of trial effect will be imprecise
Trial analysis steps
• Step 1 - to compare the characteristics of intervention and control groups at entry to the trial
–null hypothesis of `no difference between I and C groups’ – is there strong evidence
against?
Example: randomized controlled trial examining benefits of cholesterol lowering treatment (with statin) in diabetes (CARDS)
- Step 2 - to establish whether the intervention has been applied and affected relevant intermediate variables
- e.g. from trial examining benefits of statin (cholesterol lowering) treatment in diabetes (CARDS)
– Has the atorvastatin been taken?
– Has the atorvastatin led to a reduction in cholesterol levels in the intervention group?
- Step 3: does outcome differ between intervention and control groups?
- Method of analysis depends on whether outcome is continuous or categorical
– e.g. blood glucose, blood pressure (mean difference) – e.g. event/no event • Dead or alive? • Myocardial infarction/ no myocardial infarction
Trial analysis- after step 3?
- Statistical test will yield p (probability) value showing how strong evidence is against the null hypothesis of no difference between intervention and placebo…. (p < 0.05, 1 in 20 chance) conventionally accepted
- 95% confidence limits around effect estimates
• More sophisticated analyses allow adjustment for
– exact timing of events using `survival analysis (see hazard ratios rather than relative risks)
– imbalances in key factors at randomization
• Subgroup analyses (pre-specified) – are there particular participant groups who get greater (or lesser) benefit than usual?
Kaplan Meier plot
used to measure the fraction of subjects living for a certain amount of time after treatment
In clinical trials or community trials, the effect of an intervention is assessed by measuring the number of subjects survived or saved after that intervention over a period of time. The time starting from a defined point to the occurrence of a given event, for example death is called as survival time and the analysis of group data as survival analysis. This can be affected by subjects under study that are uncooperative and refused to be remained in the study or when some of the subjects may not experience the event or death before the end of the study, although they would have experienced or died if observation continued, or we lose touch with them midway in the study. We label these situations as censored observations.
The Kaplan-Meier estimate is the simplest way of computing the survival over time in spite of all these difficulties associated with subjects or situations.
DOES CHOLESTEROL LOWERING c. STATIN Rx IN TYPE 2 DIABETES REDUCE CV RISK?
Collaborative AtoRvastatin Diabetes Study (CARDS)
2838 patients 40-75 yrs with type 2 diabetes, not high cholesterol (LDL-C <= 4.14 mmol/L), no previous CVD
Randomized to 10 mg atorvastatin vs placebo
Primary outcome (ITT analysis) = cardiovascular disease (coronary events, revascularisation, stroke) after 4 years
IS IT POSSIBLE TO PREVENT THE DEVELOPMENT OF TYPE 2 DIABETES IN PEOPLE AT HIGH RISK?
- Finnish Diabetes Prevention Study
- Middle aged overweight men and women (N = 522) with impaired glucose tolerance (i.e. high risk group)
- Randomized to intervention
- weight reduction
- total and saturated fat intake reduction
- fibre intake increase
- physical activity increase
• Outcome = development of diabetes at an annual oral glucose-tolerance test (follow-up period 3.2 yrs)
WHAT INTERVENTIONS WILL HELP TO PREVENT TYPE 2 DIABETES?
Need for interventions
– reducing dietary energy intake
– increasing physical activity levels
– prevention of overweight-obesity
Trial interpretation summary
