epidemiology (lectures 9-13)

randomized control trials

• unit of analysis is the individual

• one group receives intervention

• another group doesn’t

community trails

• unit of analysis is group or community

• one community receives an intervention

• another community doesn’t

natural experiments

• unplanned events produce conditions to conduct a natural experiment

• levels of exposure differ among populations relative unaffected by other factors, so situation resembles a planned trial 

what is an RCT

• planned experiment where investigators assign participants to intervention or control condition

• test efficacy of intervention or treatment

• preventative trials: agent given to healthy or high-risk individuals to prevent disease

• therapeutic: treatment/therapy given to diseased to reduce risk of recurrence, improve survival, increase quality of life

• has high validity = gold standard in epi 

key features of RCT

• at least one intervention group, one control group

• randomize participants → equal chance of being in intervention or control

  • blinding to the highest level possible 

control group

• the comparison arm to gauge effects of treatment

• no intervention/placebo/alternative treatment

  • placebo: positive effect, think they’re getting better even when there’s no physiologic efficacy

• good control groups:

  • have the same baseline

  • are observed, monitored and followed the same way as intervention group 

• hawthorne effect: positive effect because they are being observed, change behavior

• importance

  • outline what happens if nothing was done

  • evaluate how effective/efficacious/safe the treatment is

randomization

• want intervention and control groups to look alike in all factors EXCEPT for assigned treatment

• patient and physician both don’t know which prevention/therapy is assigned

• chance is the only factor that determines group assignment

blinding/masking

• groups of randomized individuals aren’t known to anyone involved in the study

• used to avoid bias

  • selection bias

  • representative sample

  • information bias

  • outcomes

three levels of blinding

• single-blind: participants are blinded to group assignment, investigators and analysts are aware

  • protect against placebo effect

• double-blind: participants and investigators are blinded to group assignment, analysts are aware

  • protect against bias to groups from researchers

• triple blind: treatment and research aren’t obvious to participants or investigators, analyses are completed where analysts are blinded from group assignment 

conducting an RCT

1. select intervention and develop proposal

2. ethics committee review

3. assemble cohort

4. measure baseline variables

5. choose comparison groups

6. monitor participants for outcome

7. ensure compliance

selective intervention

• start with research objective

• establish if therapies are to be tested:

  • safe?

  • active against disease?

  • superior to other treatments?

  • feasible in terms of implementation?

assembling trial cohort

• inclusion criteria: population best for generalizing findings, population most efficient for answering the question

• exclusion criteria: populations that inhibit ability to control errors, population that has difficulty with therapy compliance

measuring baseline variables

• measured to characterize study cohort 

  • demographics

  • clinical factors of relevance

• final report displays baseline characteristics of intervention vs. control group

selecting outcomes

• outcome: endpoint or dependent variables

  • remission

  • stopping cancer growth

  • growth of cancer

  • mortality

• selection of best endpoint may be complicated → choose surrogate or intermediate endpoints

assuring compliance

• must monitor adherence to intervention protocol

  • self reports, direct evaluation, pill counts, metabolite levels

  • easier to comply with once-a-day treatment

    • monitor attendance and engagement (behavioral interventions)

analyzing results

• report overall risk and rates

• present comparison of risk and rates of outcomes between groups

• present relative risk of outcome in exposed vs. unexposed groups 

• measure quality of life interventions or severity of symptoms

measuring associations: risk

• risk ratio: a/(a+b)/c/(c+d)

measuring associations: rates

• rate ratio: a/person time_E/c/person time_U

ethical considerations

• chemical equipoise: balance doubt and efficacy of treatment with belief it may work

• if research know new treatment is beneficial or toxic, trail should stop

  • beneficial to control group

  • toxic to intervention group 

• Data Safety and Monitoring Boards (DSMB): monitor trial data to make determinations

advantages of RCT

• investigator controls amount, timing and frequency of exposures

• demonstrate cause-effect relationships 

• randomization and blinding ensure high validity

  • only practical approach for research questions

disadvantages of RCT

• cannot study harmful materials or methods

• interventions may not be suitable for blinding are are different from real world practice

• limited generalizability

• difficulty following people over time, miss data for outcomes

• very expensive

• hard to recruit participants

types of trials in pharmaceuticas

• non-inferiority

  • new treatment is as effective as standard of care

  • test whether treatments are similar

  • null hypothesis: treatments are different

• superiority

  • new treatment is better than standard of care

  • null hypothesis: new treatment is better than standard of care 

phases of RCT in pharmaceuticals

• phase 0: learn how drug is processed and affects the body

• phase 1: find best dose with fewest side effects

• phase 2: safety and efficacy

• phase 3: compare new drug to standard of care

• phase 4: test new drugs approved by FDA