Transcript OKU 9_chpt15

OKU 9
Chapter 15: ORTHOPAEDIC
RESEARCH
Brian E. Walczak
KEY COMPONENTS FOR THE
DEVELOPMENT OF THE CLINICIANSCIENTIST
• Significant scientific training
– 1-2 years or more….
• Protected Time
– Minimum of 30% of the time
• Adequate funding
– Sustained for a minimum of 5 years
INFERENCE
• …. The act of deriving logical conclusions from
the existing knowledge regarding a condition
• Well designed study is to
– Provide insight into the “TRUTH”
BIAS
• Nonrandom, systematic error in the design or
conduct of a study that may result in mistaken
inference about association or causation
• Types
– Recall
– Publication
– Measurement
– Selection
CONFOUNDING
• Occurs when a variable has an association
with both the independent and dependent
variable
• E.g.
– Age
– Gender
– Socioeconomic status
– Medial comorbidities
CHANCE
• The probability that two unrelated events will appear
associated by random occurrence rather than through
a causal assoication
• “Good” study need to control for chance
• Type I (alpha) error
– Truth is no association (but you think there is)
• Type II (beta) error
– Is an association (but you fail to prove it)
• Alpha = 0.05 (commonly accepted level) – could be
anything
– There is less than a 5% risk of chalking the association up
to “chance”
Power
• Probability equal to 1-beta
– Generally accepted as 0.8
– The more stringent the beta error, the narrower
the confidence interval will be and the more
certain one may be of the results in representing
the truth
STUDY DESIGN AND LEVEL OF
EVIDENCE
• Study
– Observational
•
•
•
•
•
No allocation of treatment groups
Prospective or retrospective
Descriptive
Analytic
Case reports, case series, cross-sectional study
– Experimental (not suited for determining risk factors)
• Examines the efficacy of distinct treatment options
• GOLD STANDARD:
– DOUBLE-BLIND PROSPECTIVE RANDOMIZED CLINICAL TRIAL
CASE SERIES
• Descriptive observational study
• Potential complications or successes of a
cohort (group)
• LEVEL IV evidence
CROSS-SECTIONAL SURVEY
• Observational
• Descriptive
• “Snapshot”
CASE CONTROL
• Observational
• Patients with a given outcome are compared
with patient without the outcome of interest
• RARE or UNCOMMON DZS
• Reported as “ODDS RATIO”
• Retrospective
• LEVEL III
COHORT
•
•
•
•
Observational
Relative Risk
Prospective or retro
LEVEL II or III (prospective or retrospective)
CLINICAL TRIALS
• Experimental
• Able to minimize “chance”
• LEVEL I or II
LITERATURE REVIEW
• SUMMARY OF EXPERT OPINION
– LEVEL V EVIDENCE
• Meta-analysis
– Well-organized systematic quantitative analysis of
randomized clinical trials from which one may
draw valid statistical inferences
– LEVEL I EVIDENCE
META-ANALYSIS
•
•
•
•
Quantitative analysis
Similar study designs
Must test for homogeneity
Publication bias may be assessed using a funnel plot
– Parametric (Egger’s linear regression model)
– Non-parametric (Begg’s test) methods
• Should try to control for publication bias by including
both PUBLISHED AND NONPUBLISHED STUDIES
• Summary estimate = Forest plot
DATA
• Continuous
– Numerical info
• Any given value within a range of values
• Age
• BMI
– Non-continuous variable is called “discrete”
• Ordinal
– Ordered variable (this is why it is difference than a categorical)
• Fracture Classifications
• Socio-economic status
• Categorical (nominal)
– Qualitative variables without ordering
• Gender
• Hair color
DATA DISTRIBUTION
• Continuous data
– Parametric
• Explain the distribution by a SINGLE math equation
• Gaussian distribution
– 69% of the values will fall within 1 SD of the mean
– 95% of values with fall within 2 SD of the mean
– 99% of the values will fall within 3 SD of the mean
• Mean, median, mode are all equal]
– Mean = average
– Median = “middle” value (50th %)
– Mode = the “most”
– Nonparametric
NONPARAMETRIC
• Median, mode, mean are not the same
• Right skew (positive )= Mean > median >
Mode
• Left skew (negative) = mean < median < mode
• Kurtosis = “FAT TAIL” RISK
“P” VALUE
• Probability of observation (compare this value to the alpha = level
of significance (often < 0.05)
• Not practically significant
– BUT, measures the strength of evidence in favor of the alternative
hypothesis (vs. the “null” [Ho])
• Type I error
– Concluding an association exists when in fact it occurs by chance alone
– E.g. falsely rejecting the null
• Concluding that a difference exists (potentially type I error)
• Type II error
– Concluding an associating does not exist when it really does
– E.g. falsely accepting the null
• Concluding that a difference does NOT exists (potentially type II error)
• IF no association reported, then power should be reported because
this indicates the study’s ability to actually detect a difference
DIAGNOSTIC TESTING
• Sensitivity = TP/total (TP+FN)
– 100% means that a test will ID ALL SICK PEOPLE
• A NEGATIVE RESULTS then would R/O the DZ
• Specificity = TN/total (TN+FP)
– Probability that a person without the dz will be correctly ID
• PPV = Probability that a person who’s test is + actually
has the dz
– TP/TP+FP (all positives)
• NPV = Probability that person who’s test is – actually
has no dz
– TN/FN+TN
DIAGNOSTIC TESTS
• ODDS RATIO (retrospective)
– CASE-CONTROL STUDY
– ESTIMATES RELATIVE RISK
– TP X TN/FP X FN
• RELATIVE RISK (prospective)
– COHORT STUDY
– Used to compare incidence rate in exposed and
unexposed goups