2-Evidecne For Specific Types Of Clinical Quest

Front 1

in observational studies

Back 1

-pts and providers preference determine therapy
-tx decisions (and risk of outcome) are often influenced by co-morbidities, other meds, and severity of disease, therefore
-txs are more often withheld from pts w/the poorest prognosis
-patients who are highly compliant are likely to have better outcomes, even with a worthless treatment or placebo!

Front 2

is this evidence about therapy valid?

Back 2

1. was the assignment of pts to tx randomized?
2. was the randomization concealed?
3. were the groups similar at the start?
4. was f/u of pts sufficiently long and complete? (drop out <20%)
5. were all the pts analyzed in the groups which they were randomized? (intent-to-treat analysis)
6. were pts, clinicians and study personnel kept blind?
7. were groups treated equally, apart from experimental therapy?

Front 3

example of intent-to-treat analysis

Back 3

-even if patients from the “treatment” group stop taking the medication, their results should still be included with those of the treatment group, not the control group
-ex. NSAID study

Front 4

intent-to-treat analysis

Back 4

-In most studies, not everyone who gets randomized to (or starts) treatment completes it (as planned)
-This group may be different in some ways than the group that completed the protocol
-impt to consider why

Front 5

per-protocol analysis

Back 5

-data for those who completed the study
-"as good as it gets"

Front 6

intent-to-tx analysis cont.

Back 6

-data for everyone who was randomized to tx, even if they didn't complete
-"as bad as it might be?"
-how was the end data created?
"last observation carried forward" or, predictive model

Front 7

what is the magnitude of the tx effect?

Back 7

1. Risk and Relative Risk
2. Relative Risk Reduction
3. Absolute Risk Reduction
4. Number Needed to Treat (NNT)
5. Odds and Odds Ratios

Front 8

Risk

Back 8

-number of subjects in a group who have an event divided by total number of subjects in the group
-the probability of have an event in that group = (P)

Front 9

relative risk

Back 9

-ratio of risk in exposed group to risk in non-exposed group (P1/P2)
-used in cohort studies that look at tx effects or risk factors
-can also be calculated from RCTs
-look at incidence (events over time)
**EER/ CER

Front 10

risk in group 1 = Experimental Event Rate (EER) =

Back 10

a/ a + b

Front 11

risk in group II = Control Event Rate (CER) =

Back 11

c/ c + d

Front 12

relative risk reduction (RRR)

Back 12

CER - EER / CER

Front 13

relative risk reduction disadvantages

Back 13

-doesn't reflect how common the event is w/out therapy
-doesn't discriminate b/t large and small effects
-sometimes people trying to convince you of a therapy's usefuleness use this number!

Front 14

absolute risk reduction (ARR)

Back 14

-risk in control group minus risk in treated group

ARR = CER - EER

Front 15

number needed to treat (NNT)

Back 15

= 1/ ARR
-# of pts you would need to tx with the experimental therapy to prevent one additional bad outcome, or achieve one additional favorable outcome
*typically includes time period
-look at flu example

Front 16

using NNT

Back 16

-smaller NNTs are more impressive
-NNT is balanced against seriousness of outcome we're trying to prevent
-can compare NNTs for various interventions if time intervals are different; multiply by observed time
-can "adjust" NNTs for risk of bad outcome in our pt compared to average risk in study pts

Front 17

Odds ratio (OR)

Back 17

-the odds of an event are calculated as the # of events divided by the # of non-events
-the odds of an event in a pt in the experimental group relative to that of a pt in the control group
ex. on average 51 boys are born in every 100 births, so the odds of any randomly chosen delivery being that of a boy is 51/49 = 1.04

Front 18

odds ratio calculation

Back 18

-Ratio of odds for exposed group divided by odds for not exposed group
P1/ (1 - P1) (exposed) / P2/ (1 - P2) (not exposed)
-if # is <1 than it usually means the ppl in the group did better)

Front 19

when events are rare, risks and odds...

Back 19

are very similar!

Front 20

when are ORs used?

Back 20

-in case-control studies where "exposures" are measured, and exposed group has worse outcome
-when looking for harm expect OR > 1
-in systematic reviews, b/c of mathematical properties ( 0 to infinity)

Front 21

odds ratio and clinical trials

Back 21

-Clinical trials typically look for treatments which reduce event rates, so when looking for benefit, expect OR less than one (one = no difference)

Front 22

how precise is the estimate of the tx effect?

Back 22

-the point estimate of a study sample's OR or RR must be interpreted in light of the 95% confidence interval (CI)
-if the CI includes 1, then the true value for the population's OR or RR could be 1 (meaning no difference b/t control and tx groups), which means the results are not statistically significant!
-if it doesnt include 1 then the results are statistically sig

Front 23

systematic review

Back 23

-“A summary of the medical literature that uses explicit methods to systematically search, critically appraise, and synthesize the world literature on a specific issue”.
-goal to minimize: bias and random error

Front 24

meta-analysis

Back 24

-a systemic review that uses quantitative statistical methods to summarize the results of research (combining the results of individual studies), preferably randomized controlled trials

Front 25

example: For studies measuring pain at 2-7 days...

Back 25

(look at example!)
1. which of the studies showed pain was better with abx?
2. what is the CI for the Thalin study is it significant?
3. Which study has the most precise CI?
4. what is the CI for the combined studies? Is it significant?
-look at math slide 46

Front 26

Clinical guidelines, the good, the bad and the ugly!

Back 26

1. provide guidance for typical pts with a common condition
2. variable research bias
3. sometimes various groups interested in a subject provide conflicting guidelines!

Front 27

components of guidelines

Back 27

1. Evidence Summary: systematic review of all relevant world-wide literature
2. Detailed instructions for applying the evidence to our pt

Front 28

Guides for deciding whether a guideline is valid?

Back 28

1. Did its developers carry out a comprehensive, reproducible literature review within the past 12 months?
2. Is each of its recommendations both tagged by the level of evidence upon which it is based and linked to a specific citation?

Front 29

guidelines- applicable to my pt/practice/hospital community?

Back 29

1 is the BURDEN of illness too low to warrant implementation?
2. Are the BELIEFS if individ. pts or communities about the value of the interventions imcompatible with the guidelines?
3. would the opportunity cost of implementing this guideline constitute a bad BARGAIN in the use of our E or our community/s resources?
4. Are the BARRIERS so high that it is not worth trying to overcome them?

Front 30

Types of studies on Harm/Etiology

Back 30

-not too likely to find reviews of RCTs related to harm
-often have to rely on other types of studes
- Cohort, Case-control, cross-sectional studies, & case reports or case series

Front 31

Harm/Etiology: validity

Back 31

1. Were there clearly defined groups of patients, similar in all important ways other than exposure to the treatment or other cause?
2. Were treatments/exposures and clinical outcomes measured in the same way in both groups?
3. Was the follow-up of the study patients sufficiently long (for the outcome to occur) and complete
4. Does it seem likely that the exposure CAUSED the harm?

Front 32

Is this valid evidence about harm impt?

Back 32

1. Relative Risk (RCTs and cohort studies)
2. Odds Ratio (Case-Control studies)
3. Number Needed to Harm (NNH)

Front 33

Relative risk in a 1. randomized trial or cohort study and in a 2. case-control study

Back 33

1. [a/(a + b)]/ [c/c + d)]

2. ad/bc

Front 34

Number needed to harm (NNH)

Back 34

-a more clinically useful number
-NNH = # of pts you would need to expose to a harmful entity to produce one additional harmful outcome
-for RCT an cohort studies, NNH = 1/RR
-for case-control studies, NNH is based on patient expected event rate (PEER), which is the event rate in the un-exposed group (more complicated)