Transcript Understanding Medical Articles and Reports

Understanding Medical
Articles and Reports
Linda Vincent, MPH
UCSF Breast SPORE Advocate
September 24, 2005
[email protected]
510-527-6720
Introduction to Statistics
 Mean, average of a set of numbers or values
 Median, the number closest to the middle of a set
of numbers
 Standard Deviation, information on the spread of
the numbers from a complex calculation
More Statistics
 Probability, p-value, comparing sets of numbers from
research in another complicated calculation that gives
information on statistical significance
 p<.05 is the same as 95% Confidence Intervals
 Answers question like, how likely is it that
people in the study
that got the investigational drug got better because the drug did its
job, or whether they got better due to some other factor.
Risk Ratio
 Relative Risk or Risk Ratio, RR is used widely in research
papers. Compares the probability of an outcome among
individuals who have a specific “exposure” or a specific
characteristic to those who have not been exposed
 RR is calculated in Cohort or Experimental studies (Clinical
Trials)
Calculating Relative Risk
2 X 2 Table, see page 51 of Cook Article, Lung
Number with
disease
Number
without
disease
Number of
exposed
90
19,846
Number of
unexposed
115
19,827
Calculating Relative Risk-2
 RR Calculation:
RR = Absolute risk in exposed participants
Absolute risk in unexposed participants
90/19,934
115/19,942
RR = 0.78
CI = 0.59-1.03
p-value = .08
Interpretation of RR
 RR equal to or greater than one, written as =>1 The
higher the number above one the greater the
risk/disease is associated with the exposure
 RR =< 1 There is no association of the risk/disease
to the exposure
 RR = 1 The risk is the same in both groups
Interpretation of Confidence Intervals
 Confidence Intervals, CI
 95% CI is current standard
 If the CI does not include 1.0 that the RR is statistically
significant
 Previous example in the Cook article, CI = 0.59-1.03
Odds Ratios
Case-Control Studies can’t use RR because the OR, Odds
Ratios are an estimate of RR
Number with
disease (cases)
Number exposed
Number not
exposed
Number without
disease (controls
Absolute Risk
 Absolute Risk gives an actual number of health
problems that happened or were prevented due to a
drug
 Doctors and patients have difficulty understanding
RR and OR
Use of NSAIDS and Risk of Breast Cancer: The CaseControl Surveillance Study Revisited
 Advantages of Case-Control Studies:
 Less expensive and time-consuming than cohort
studies especially with rare diseases
 Odds Ratio can be used to estimate the Risk Ratio
Use of NSAIDS and Risk of Breast Cancer: The CaseControl Surveillance Study Revisited
 Disadvantages of Case-Control Studies:
 Cases and Controls may come from different groups which
will bias the results
 Time-related relationships may be difficult to ascertain
 There is the always the possibility of recall bias when asking
about exposures that occurred in the past
Use of NSAIDS and Risk of Breast Cancer: The CaseControl Surveillance Study Revisited
 What is the purpose of the study and what question did the
authors address?
 See first sentences of the abstract
 What were the unanswered questions raised by prior
scientific work that led to this study?
 The meta-analysis of Khuder and Mutgi found an 18%
decreased risk in regular NSAIDs users in 8 case-control
studies.
Critical Evaluation of Zhang Article, Continued
 Do the authors state a hypothesis? If so what is it?
 Hypothesis is a statement of belief used in the evaluation of
population values.
 Hypothesis statement: There is an association between the
exposure and the outcome.
 No hypothesis statement in the Zhang article
Zhang, continued
What are the results?
Table 1-For the most relevant category, regular use of
NSAIDs, OR was 0.78, 95% Confidence Intervals
(CI) 0.63-0.97
 CI dosen’t include 1.0
Table 2-Longer duration of regular NSAIDs use was
associated with a lower risk of breast cancer, pvalue for trend was 0.02
Table 3-Looked at differences between aspirin,
Ibuprofen, and NSAIDs (p-value for trend, not
significant)
Zhang, continued
 What are the conclusions drawn from the results? Are the
conclusions supported by evidence in the article?
 First paragraph under DISCUSSION, “Our results suggest a
modest decrease in the risk of breast cancer associated with
regular NSAIDs use….”
 Last paragraph, “In summary, the present study found…the
effect is relatively weak. Given the conflicting results from
large cohort studies…too early to suggest that regular use of
NSAIDs could help prevent breat cancer.”
Zhang, continued
 What methods did the authors use?
 Cases were selected from patients living within one
hour driving distance to the hospitals
 Cases were interviewed by trained nurse interviewers
using structured questionnaires
 Controls…
 Data analysis, relationships of use to risk performed
with multiple logistic regression models
Zhang, continued
 What are the strengths of the paper?
 Researchers/authors attempted to validate previous
studies on use of NSAIDs and reducing risk of
breast cancer
 Note on biological mechanism for reducing risk
 Note on inconsistencies from other studies
Zhang, continued
 Are there obvious problems with the article?
 The hormone receptor status of the tumors was obtained from
only 26% of the cases. Other risk factors for breast cancer were
not evaluated, but were similar among cases with data on
hormone status suggesting absence of selection bias
What additional experiments would strenghten the authors
conclusions?
Are the findings significant? If so, for who?
Resources
 Free subscription services




BreastCancer.Net, http://breastcancer.net
Amedeo.com, http://amedeo.com
Medscape, http://www.medscape.com
PubCrawler, http://pubcrawler.gen.tcd.ie
Resources continued…
National Cancer Institute
Pub Med, http://pubmed.gov
MedLinePlus, http://medlineplus.gov
FDA Center for Drug Evaluation and Research
http://www.fda.gov/cder
Guidelines
National Comprehensive Cancer Network,
http://nccn.org
San Antonio Breast Cancer Symposium
http://sabcs.org
American Society of Clinical Oncology
http://asco.org
Project LEAD
 http://natlbcc.org or Google National Breast Cancer
Coalition
 Check left column links, click on Education and
Training, and click again on Project LEAD