Purpose of the lecture.
To describe the dual role of quantitative assessment of evidence in medicine‹applying the evidence to patient care and acquiring new medical knowledge, and to introduce the standard approaches to clinical investigation, including clinical trials, case-control studies, and cohort studies, and to examine their relative strengths and weaknesses.

Student goals for the lecture.
- Understand the scope of epidemiology and biostatistics
- Understand the role of quantitative methods in medicine
- Know difference between observational studies and experiments

Lecture notes

€ Evidence-Based Medicine

• Effective medicine is better than ineffective medicine
  
• Treatments that have strong evidence of effectiveness are better than those that don't
  
• Evidence comes from:

• Not all patients respond to treatment the same way
  
• Treatments that are effective for my patients are better than treatments that are generally effective
  

A Sources for continuing medical education: journals, pharmaceutical mfrs, meetings.
B Generating new knowledge: experiment and observation
C Dissemination: the medical literature
D Personal acquisition: triage of medical information

II Primary problems of epidemiology and biostatistics

A Epidemiology: "Study of health and illness in human populations"
1 General: what factors effect patterns of health and disease in human populations?
2 Clinical epidemiology: Applying population characteristics to mgmt of individual patients
Sources: CDC, Regular JAMA reports
3 From Sackett, et al: "Three challenges:
a) Reaching the correct diagnosis
b) Selecting management that does more good than harm
c) Keeping up to date with useful advances in medicine

B Statistics: Coping with variability in a scientific way
1 Questions.
When something is contrary to expectation, is it real or due to inherent variability?
By how much do two therapies differ?
What range of effects are consistent with the data we have observed?
Is a particular patient's symptoms "unusual" or "normal"?
Is what we see signal or noise?
2 Statistical inferences (summaries of evidence) are based on
a) the comparison of
b) an observed effect with
c) a measure of underlying variability.
3 Thus, statistics consists of
a) Measures (=descriptions) of the size of effects based on observed data
b) Measures (=descriptions) of inherent variability
c) Methods (=standard techniques) for making "noisy comparisons"
4 Makes it possible to draw inferences of general applicability
Apply knowledge derived from many patients to individual clinical problems

III Quantitative Methods in Medicine

A Medical literature cannot be accepted uncritically
1 Physicians depend on med lit to stay current
2 Conclusions depend critically on manner in which data are collected and combined
3 80% of literature reports have inadequate design, execution, or analysis.

B Reader must determine for self validity and applicability of reported findings

C Applying study results to patient care requires understanding of stats
Example. Elimination of benzodiazapenes
- Indicated for anxiety. Example: before surgery
- PDR: 2­20 mg IV ("Lower doses, usually 2­5 mg, should be used for elderly pts")
- Duration of action of some importance to anesthesiologist
- Diazepam (Valium) and midazolam (Versed) are two widely used anti-anxiety drugs
- Midazolam 1/2 life increases slightly with age; marked increase (20Æ75 hrs) for diazepam
- „50 yrs, there is more variability between patients for diazepam elim than for midazolam

• Understanding this pattern of variation leads to the following conclusions for pt care:

D Other examples:
1 Interpreting vital statistics
2 Using epidemiological information
a) Disease prevalence
b) Incidence of complications (eg, death secondary to lumbar spine surgery)
c) Assessing risk factors (smoking for pulmonary, cardiovascular disease)
3 Correct interpretation of diagnostic procedures: interpretation of abnormal values

IV Study designs in medical research
Study design: the procedure under which a study is carried out
A Two main categories: Observation and experiment
Observation: Identify subjects and then observe and record characteristics
Experiment: Identify subjects, place in common context, observe effects of intervention
1 Observations are readily obtained, but subject to bias, that is systematic errors
2 Some observational designs are less subject to bias than others
3 Experiments are hard to do well
4 Experiments can answer narrow questions definitively
5 Generalizability of results from experiments may be at issue, eg new drug testing that excludes women subjects

B Observational studies
A control serves as a standard for comparison, both for effects and for variability.
1 Case series, eg, "Normal plasma cholesterol in an 88-year-old man who eats 25 eggs a day"
[Kern J, NEJM 1991; 324:896­899]
a) Excellent at identifying unusual situation
b) Good for generating hypotheses amenable to rigorous test
c) Generally short-term, investigators self-select (bias!), generally no controls
2 Case-control studies, eg, "Diagnostic X-ray procedures and risk of leukemia, lymphoma, and multiple myeloma," [Boice JD, et al, JAMA 1991; 265: 1290­1294]
a) Controlled studies; retrospective.
b) Identify cases (with condition of interest, here, leukemia); match to disease-free controls who are similar with respect to known risk factors for condition; compare degree of exposure to possible risk factor (here, diagnostic xray).
1) Logic: The cases differ from the controls only in having the disease. If the exposure does not in fact predispose to having the disease, then the exposure should be equally distributed between the cases and controls. The extent of greater previous exposure among the cases reflects the increased risk that exposure confers.
2) Measure of effect: Either the relative risk (ratio of probabilities of contracting disease given exposure) or the odds ratio (ratio of the odds of contracting disease given exposure). RR or OR of 1 indicate no effect of exposure (equal odds). ["Patients with non-Hodgkin's lymphoma were exposed to diagnostic x-ray procedures more often than controls (RR, 1.32), but the RR fell to 0.99 when the exposure to diagnostic x-ray procedures within 2 years of diagnosis was ignored."]
3 Cohort studies
a) Prospective
b) Can determine causes and incidence of diseases as well as identify risk factors
c) Generally expensive and difficult to carry out
d) Identify groups of exposed subjects and control subjects according to exposure. Match for other risk factors. Follow over time, record in each group the fraction who develop the condition of interest, and compare these fractions.
1) Measure of effect: RR or OR
2) Logic: Differences in the rate at which exposed and control subjects contract a disease is due to the differences in exposure, since other known risk factors are equally present in the two groups.

C Clinical Trials: Experimental studies
Clinical trial: A comparative, prospective experiment conducted in human subjects. (4 reqs)
1 Controlled vs uncontrolled studies
a) Controlled Æ comparison made relative to a simultaneous reference condition
b) UncontrolledÆ comparison implicit (previous experience, historical evidence, anecdote)
c) Controlled trials make it possible to ascribe differences in outcome to differences in treatment; differences from expectation in uncontrolled studies could also be due to many other factors (different kinds of patients, change in standard of care, improving technique, etc)
2 Concurrent controls: randomized vs nonrandomized assignment to Rx
a) "The randomized clinical trial is the epitome of all research designs because it provides the strongest evidence for concluding causation."
b) Conclusions from nonrandomized studies subject to many sources of bias
c) Patients can serve as own controls; this is usually beneficial, as the comparison removes patient differences.
1) Example: "Effects of Vasocon-A in the allergen challenge model of acute allergic conjunctivitis" [Ableson MB et al, Arch Ophthalmol 1990; 108:520­524]. Subjects received placebo eye drops in one eye, Vasocon-A drops in the other eye. (Randomization used to determine which eye gets placebo for each patient.) Cat-dander extract applied to both eyes, response between eyes compared.
2) Paired designs work best when outcome can be observed shortly after treatment, and disease/treatment short-lived
3) Crossover study is a special case: each individual receives each treatment for a period of time and the responses during each period are compared. Many practical difficulties with cross-over study (eg, possible carry-over effects)
3 Blinding: Good practice dictates that factors that can affect the evaluation of outcome not be not be permitted to influence the evaluation process.
a) Double-blind design: neither patient nor outcome evaluator (usually, the investigator) knows the treatment to which the patient was assigned when the evaluation was made.
b) Single-blind: Patient or evaluator is blinded as to Rx, but not both.
c) Triple-blind: Pt, Phys, and Data analyst are blinded as to Rx identity
4 Historical controls are better than no controls, but not by much. Examples: ddI and AZT for AIDS; gastric freezing for peptic ulcer.

References:

Sackett DL, Haynes RB, Guyatt GH, Tugwell P Clinical Epidemiology: A Basic Science for Clinical Medicine. Second edition. Little, Brown: Boston, 1991.