TREATMENT 1Evaluation of interventions

How best assess treatments /other interventions?

RCT (randomised controlled trial)

OBJECTIVES

Treatment lecture 1 Describe structure of RCT Define, calculate and interpret main measures of

effect for RCTs Compare RCT design with observational study

designs

(Treatment lecture 2) Explain differences between efficacy and

effectiveness Distinguish between explanatory and

management trials)

What will be covered

Experimental design RCT architecture

Architecture of cohort and case control studies RCT analysis (measures of effect)

Calculations Examples……….

RCT is Main tool for assessment of treatments

/other interventions Gold standard for treatment evaluation

RCT design based on experiment

What is an experiment? See if you name the defining characteristic(s) of

an experiment

Experiment1. Intervention group identical to control group2. Random allocation of study factor (by researcher)3. All other factors constant

Hence any differences in outcome between the study and control groups can only be due to the intervention i.e. the study has proved that the intervention caused/prevented the outcome.

Not possible to fully achieve these characteristics in research on humans. A randomised controlled trial (RCT) is as close as we can get i.e. a quasi experimental design.

In an RCT1. random selection 2. random allocation of factor3. double blind proceduresare used to approximate the characteristics of an experiment.

Randomised controlled trial

population

group 1

group 2

Outcome?

Outcome?

new treatment

control treatment

Time

How assess impact of harmful factors e.g. alcohol, smoking, radiation?

RCT? Not ethical to assign one group to interventions

that are thought to be harmful! Observational studies

Cohort study Case control study Ecological / correlation study

(Observational epidemiology)

Cohort study

population

group 1

group 2

Outcome?

Outcome?

exposed

not exposed

self selected

Time

Case control study

population

% Exposed?

% Exposed?

Case(Has disease X)

Control(Not disease X)

Time

Case control study

population

% Exposed?

% Exposed?

Outcome(disease X)

Not outcome(Not disease X)

Time

RCT architecture

Total patient population(reference population)

Total number of patients in trial

Treated Placebo

No. (%) of outcomese.g. deaths, cures

No. (%) of outcomes,e.g. deaths, cures

RANDOMISATION

Informed consent

Analyses for RCTs

How compare outcomes in treated and control groups? Example 1:

2% mortality rate in treated (Rx)

vs 4% in controls?

The obvious way to compare these two proportions is to either Divide Subtract

Example 1: 2% mortality rate in Rx vs 4% in controls?

Divide: 2% / 4% = 0.5 Half as many deaths in treated groupTreatment is better than controlIrx / Ic = 0.5 = relative risk (RR)

Or Subtract: 2% - 4% = - 2% (minus)

2 fewer deaths in treated group for every 100 treatedTreatment is better than control

Irx - Ic = - 2% = risk difference (RD)Ic - Irx = 2% = risk reduction (RRed)

Number needed to treat (NNT) to prevent 1 death = 1/risk reduction = 1 / 2% = 50

Example 2: 2/10,000 mortality rate in Rx vs 4/10,000 in controls?

2/10,000 / 4/10,000 = 0.5 Half as many deaths in treated groupTreatment is better than controlIrx / Ic = 0.5 = relative risk (RR)

Or 2/10,000 - 4/10,000 = - 2/10,000

2 fewer deaths in treated group for every 10,000 treated!!Treatment is better than control - v. few deaths prevented!

Irx - Ic = - 2/10,000 = risk difference (RD)Ic - Irx = 2/10,000 = risk reduction (RRed)

NNT = 1 / 2/10,000 = 500

Note that relative risk in Examples 1 is the same as in Example 1 but that the risk difference, risk reduction and NNT are very different in the two examples. Outcome measures that based on subtraction are dependent on the magnitude of the rates (per 100 vs per 1000 etc.) whereas the magnitude is cancelled out in relative rates.

Hypothetical mortality rates at 10 years

Outcome

Yes No Incidence

Treatment

n = 501 49 2%

Controln = 50

2 48 4%

RR =

OR =

RD =

RRed =

RRR =

NNT =

Try to calculate these outcome measures yourself before turning to next slide

Hypothetical mortality rates at 10 years

Outcome

Yes No Incidence

Treatment

n = 501 49 2%

Controln = 50

2 48 4%

RR = 0.5

OR = 0.49

RD = -2%

RRed = 2%

RRR = 50%

NNT = 50

Hypothetical mortality rates at 10 years

Outcome

Yes No Incidence

Treatment

n = 501

(a)

49

(b)

2%

Controln = 50

2

(c)

48

(d)

4%

RR = 0.5

OR = 0.49

RD = -2%

RRed = 2%

RRR = 50%

NNT = 50

OutcomeYes No Total

Exposed a b a+b

Not exposed c d c+d

Total a+c b+d a+b+c+d

Incidence(E) = a/a+b Incidence (NE) = c/c+d

RR = Inc(E) / Inc(NE) = a/a+b / c/c+d

OR = a/b / c/d = ad/bc

RRed = Inc(NE) - Inc(E) = c/c+d - a/a+b

RRR = RRed (x100) Inc(NE)

NNT = 1/ RRed

Formula based definitions of outcome measures

Relative risk (RR) = incidence in treated group incidence in control group

Odds ratio = Outcome/ no outcome in treated group Outcome/ no outcome in control group

Risk difference (RD) (Attributable risk)= (incid. in treated group) - (incid. in control

group)(Absolute) Risk reduction (RRed)

= (incid. in control group) - (incid. in treated group)

Relative risk reduction (RRR) (%) = risk reduction (x100) incidence in control group= 1 - RR (x 100)

Number needed to treat (NNT) = 1/ risk reduction

Text based definitions of outcome measures

Treatment (exposure) Bad outcome eg death

Rx good Rx bad(protective)

<1 1 >1Relative risk

Rx bad Rx good

Good outcome eg cure

Interpretation of relative risk values

Sustained virological response(48 wks)O. Reichard et al. Lancet 1998, 351,83-7Randomised, double-blind, placebo-controlled trial of interferon alpha-2b with and without ribavirin for chronic hepatitis C

OutcomeYes No

Incidence RR = CI.95 1 - 4.0P = 0.7OR = RD = RRed = RRR =NNT =

Treatmentn=50

18 32 ?

Controln=5

9 41 ?

Calculate the outcome measures for this study

Note that a sustained virological response is a good outcome so if the Rx works, we should have higher incidence in the treated group

Analyses for RCTs

How compare outcomes in Rx and control groups?

Example 3:

9% cure rate in Rx

vs 3% cure rate in controls?

Divide Subtract

Example 3:

9% / 3% = 3.0 3x as many cured in treated group as in

control Treatment is better than control Irx / Ic = 3.0 = relative risk (RR)

Or 9% - 3% = (+)6%

6 extra cures in treated group for every 100 treated

Treatment is better than control Irx - Ic = 6% = risk difference (RD) (risk reduction not relevant)

Treatment (exposure) Bad outcome eg death

Rx good Rx bad(protective)

<1 1 >1Relative risk

Rx bad Rx good

Good outcome eg cure

Interpretation of relative risk values

Sustained virological response(48 wks)O. Reichard et al. Lancet 1998, 351,83-7Randomised, double-blind, placebo-controlled trial of interferon alpha-2b with and without ribavirin for chronic hepatitis C

OutcomeYes No

Incidence RR = 2 CI.95 1 - 4.0 P = 0.7OR = 2.6RD = 18%RRed = -18%RRR = -100%NNT = 5.6

Treatmentn=50

18 32 18/50=36%

Controln=50

9 41 9/50=18%