Title: Endpoints in clinical studies
1Endpoints in clinical studies
- Mark Conaway
- Div of Biostatistics and Epidemiology
2Outline
- Definitions
- Classification of endpoints
- Multiple endpoints
- Surrogate endpoints
3Endpoints
- Quantitative measurements required by or implied
by the objectives of the study/trial - Desirable features of endpoints
- Relevant to disease process, easy to interpret
- Free from measurement or assessment error
- Sensitive to treatment differences
- Measurable within a reasonable period of time
4Prefer hard endpoints
- hard endpoints clinical landmarks that are
- well-defined in study protocol
- definitive with respect to disease process
- not subjective
- Examples
- death,
- time to disease progression/relapse,
- some laboratory measurements
5Soft endpoints
- Soft endpoints
- not directly related to disease process or
- require subjective assessment by
patient/physician - Examples
- Quality of life questionnaires
- Symptom questionnaires
6Not all endpoints can be classified
- Some endpoints are useful and reliable but
require some subjectivity - Examples
- pathology
- Key issue is not the classification as hard or
soft, but how prone to error is the endpoint ?
7Multiple endpoints and type I error
- Often a discussion with regard to clinical trials
but applies to observational studies as well - Randomly assign n patients to group A and n
patients to group B. - Or have well-defined cohorts of patients treated
8In principle
- Have chosen an endpoint
- Clinical relevant
- Have an appropriate sample size
- to have a type I error rate of 5
- sufficient power for a clinically meaningful
difference
9In practice
- Rare that trials use a single endpoint
- Endpoints
- cover clinical events
- symptoms
- physiologic measures
- side effects
- quality of life
10Example
- Example CALGB 9182
- Survival time
- Time to disease progression
- Response
- PSA (at 8 week intervals)
- Quality of life (5 instruments at 8 week
intervals) - Toxicity (90 item checklist)
11Example Michalowicz et al (NEJM, Nov 2, 2006)
- Study of periodontal therapy and birth outcome
- Several outcomes of interest
- preterm birth (before 37 weeks),
- birth weight,
- proportion of infants who are small for
gestational age, - Apgar scores,
- admissions to NICU
- .
12Whats the problem?
- If test each endpoint at the 5 level
- overall chance of finding at least one endpoint
where there is a significant difference is larger
than 5, even if the treatments are identical - Prone to distorted reporting (i.e. pick most
significant) - Good reference Pocock (1997) Controlled Clinical
Trials, p 530-545
13Dealing with problems with multiple endpoints
- Have a pre-defined strategy
- Some advocate
- all results pre-written, with results filled in
as trial concludes - Alternative view
- need to be flexible
- need to allow for unexpected findings
- but recognize potential for problems type I
error rate is not 5
14Delineate primary and secondary outcomes
- Many advocate having a single primary endpoint
- drives sample size calculations
- test based on this endpoint has a 5 type I error
rate - All other endpoints are secondary
15Example Michalowicz et al (NEJM, Nov 2, 2006)
- Primary
- Gestational age at end of pregnancy
- Secondary
- birth weight, proportion of infants who are small
for gestational age, Apgar scores, admissions to
NICU
16Delineate primary and secondary outcomes
- Can be hard to adhere to in practice
- For example, what if primary outcome is not
different among groups, but all secondary
outcomes are?
17What to do with primary and secondary endpoints?
- ONeill, R. (1997) Secondary endpoints cannot be
validly analyzed if the primary endpoint does not
demonstrate clear statistical significance
Controlled Clinical Trials, 550 556 - Davis, C.E. (1997) Secondary endpoints can be
validly analyzed, even if the primary endpoint
does not provide clear statistical significance
Controlled Clinical Trials, 557 - 560
18ONeill (1997)
- Primary endpoint definition
- clinical endpoint that provides evidence
sufficient to fully categorize clinically the
effect of a treatment that would support a
regulatory claim for the treatment - Secondary endpoint
- additional clinical characterization of a
treatment but could not, by itself, be convincing
of a clinically significant treatment effect
19ONeill (1997)
- Argues that primary and secondary outcomes are
generally related - Analysis of secondary should be conditional on
the primary outcome analysis result - especially true when secondary outcomes depend
directly on primary (survival) - Cant quantify the uncertainty in analyses done
after looking at results
20Davis (1997)
- Strict adherence could miss important and
unexpected results - Argues that the major problem is multiple
comparison issue - its a statistical problem, so use a statistical
solution - One such solution is the Bonferroni adjustment
21Bonferroni procedure
- If you have k endpoints
- Multiply observed p-value by number of endpoints
- For example, with k 8, convert an observed
p-value of 0.01 to 0.08 - Ensures that if Ho is true for all endpoints,
probability of rejecting Ho for at least one
endpoint is less than or equal to ?
22Limitations forBonferroni procedure
- Endpoints tends to be correlated, so this is
conservative - probability of type I error is much smaller than
? - Treats all outcomes as equal in importance
- Can lead to illogical results
- Trial with p-values 0.01, 0.75,0.75,0.75,0.75
significant - Trial with p-values 0.02, 0.02, 0.02, 0.02, 0.02
is not significant
23Limitations forBonferroni procedure
- Procedure reduces power to detect real
differences in specific outcomes, if they exist - Protect type I error at expense of power
- Difficult to apply strictly in many cases
24So whats the answer?One persons opinion...
- Selection of a primary outcome is important
- Need to allow for surprises
- Full disclosure of endpoints, instead of selected
endpoints, can alleviate a lot of the problems - Adjust p-values?
- Whats the goal of the study?
25Surrogate endpoints
- Hesitate to use the term
- Has a specific technical definition
- Issue
- Quicker, less expensive, less clinically relevant
endpoint or - More expensive, clinically definitive endpoint?
26Example
- Treatment for osteoporosis
- Endpoint
- Bone density via DEXA?
- Fracture?
- If fracture, how would this be ascertained?
27Example
- Choice is, for same amount of resources
- more patients with less clinically relevant
outcome (bone density) - Fewer patients with more clinically relevant
outcome (fracture) - Frequently see the quick endpoint in earlier
stage trials.
28Summary
- Choice of endpoints is crucial to the success of
the study either RCT or observational - Issues about
- Which one?
- How many?
- Primary vs secondary?