Observational Studies

Why the caution?
´Dr Rao found that patients who had had a transfusion because of a low red blood-cell count had an 8% chance of dying within 30 days. Without a transfusion, only 3% died. Those numbers need to be treated with caution.µ
The Economist Magazine, 2007

.Chance. Rao found an association with blood transfusion. or just chance? If not chance. or is something else going on? Dr. but he cautioned that the effect of severity of illness might have gotten mixed up with the effect of the exposure. Is this a real difference.3% = 5% more deaths within 30 days with blood transfusion. what else might be going on? Dr. Rao found a difference of 8% .

Confounding variables Dr. death . Rao·s study of patients· survival after blood transfusion is observational. for the sound of two feet dropping« One: severity of illness correlates with the exposure. and the severity of illness is a confounding factor« Definition Listen. blood transfusion Two: severity of illness correlates with the outcome.

True or false. but alcohol consumption is a confounding variable. . and explain« Some studies find an association between liver cancer and smoking.

Matching Three levels of severity Low ´Exposedµ ´Controlµ Medium High II.Two strategies« I. Randomized clinical trial Treatment Eligible Pop·n or the ´pop·n at risk«µ 1/2 1/2 Control .

double-blind experiment Clofibrate Number Placebo Number 1. controlled.789 Deaths 15% 28% 21% Deaths 15% Adherers Non-adherers 708 357 1.103 25% 20% Total There are two conclusions here. .The clofibrate trial for heart disease prevention« a randomized.813 882 2.

Experiments and Observational Studies Method of comparison  T versus C or Exposed versus Unexposed Experiments  Key variables are held constant  Control over group assignment. called randomization. not interventions  Key variables may vary. where« «impartial chance. works best  Treatments are interventions. Observational studies  No control over group assignment. strategic but not always effective . often« «self selection or exposure by association«  Observations. so« «confounding is always a risk  Matching. where« « the goal is a description of the effect.

$ and happiness in Switzerland Modes of scientific discovery  notebooks. monetary policy)  natural experiments.g. welfare) and economics (e.Why not always experiment? Ethics and human experimentation  smoking and lung cancer Prohibitive and costly  social policy (e. laboratories. and theories .g.

For men and for women in each age group. but once you·ve started. don·t stop. those who had never smoked were on average somewhat healthier than current smokers. (a) (b) Why study men and women and the different age groups separately? The lesson seems to be that you shouldn·t start smoking. . Comment. but the current smokers were on average much healthier than those who had recently stopped smoking.Smoking and health The Public Health Service studied the effects of smoking on health in a large sample of representative households.

Cervical cancer and circumcision« Cervical cancer was once quite common. In the 1950s. Epidemiologists trying to identify the cause had noticed that in several different countries it was rare among Jews and Moslems. Was this conclusion justified? Discussion *What is the response or outcome? the ´exposureµ or treatment? * Do these communities differ from members of other communities in ways besides circumcision? . some investigators concluded that circumcision of males was protective.

even after adjusting for age. California. users of oral contraceptives have a higher rate of cervical cancer than non-users. 462-69«. explain.what other factor? Were the conclusions justified? Yes or no. and marital status. 106.* Experiment? Or. *American Journal of Epidemiology. Investigators concluded that the pill causes cervical cancer. 1977. number of PAP smears before entry. education.adjustments were made for religion. and ´selected infections. smoking. Vol.µ .Cervical cancer«1970·s According to a study done at Kaiser Permanente in Walnut Creek. pp. observational study? Why the adjustments for«? Users were likely to differ from non-users on another factor affecting the risk«.

They can even be spurious. . but they are not causal.What explains the observations? What is the human papilloma virus? And how is it spread? Associations establish links.

True or false. according to many observational studies. experiments could easily have reached the wrong conclusion. or b. These studies were so encouraging that two randomized controlled experiments were done«. Experiments confirm the results of the observational studies. In what way might the observational studies have gotten it right? . Findings were« 1. 2. 2. No difference in death rate for colon cancer between T and C. observational studies could easily have reached the wrong conclusion.Vitamins«associations and effects« People who get lots of vitamins by eating five or more servings of fresh fruits and vegetables (especially ´cruciferousµ vegetables like broccoli) have much lower death rates from colon cancer and lung cancer. People who eat lots of fruits and vegetables have lifestyles that are different in many other ways too³so due to confounding the a. Beta carotene (as a dietary supplement) increased death rates due to lung cancer. and explain 1.

however. Rao had established an association or link between blood transfusion and survival. Nurses Health Study found a link between hormones and cardiovascular disease.Associations establish links« Here are two examples« Dr. It did point to something causal: if transfusion was a cause of death. WHI did. corroborate the risks« . It did point to something causal: if hormones are protective. then you·d expect«.but this alone does not prove causation. then you·d expect«but the Women·s Health Initiative did not confirm the protective effects in an randomize controlled experiment.

³while introducing the vaccine to the treatment group but not the controls. Randomization made the controls like the treatment group in important respects. HIP Screening Trials held certain key variables constant. blinding. except for the intervention. . Randomization made«. while introducing screening to treatment and not the controls. so that the difference in response is likely to be due to the effect of the vaccine itself.so the difference in death rates is likely to be due to the lives saved by screening and early detection.Experiments measure effects« Here are two examples Salk Vaccine Field Trial held certain key variables constant³ protocol. etc.

4 oz. Randomization and the response  Amount Nitrogen assigned randomly to plots  Ounces of rice. Response schedule states the effect predicted yield = (20 oz rice per oz nitrogen ) × nitrogen + 240 oz .International Rice Research Institute. or yield. moisture. 16 oz  light. Philippines I. and key nutrients held constant II. 12 oz. 8 oz. grade. is the response III. Subjects and the treatment Subjects: 20 experimental plots planted with IR 8 Treatment: 5 amounts of Nitrogen fertilizer  0 oz.

experiments measures the effects of interventions.. the investigators do not assign subjects to treatment or control«as in the case of smoker versus non-smokers. when things go as planned.Summary 1. Study design is a central issue in applied statistics. With observational studies. Why the caution? Confounding factors« 3. Observational studies can establish associations. 2. 4. In observational studies.. then the exposed ought to be sicker than the unexposed. The great weakness of observational studies is confounding. try to find out how the subjects came to be in treatment or in control. randomized controlled experiments minimize this weakness. as shown here by contrasting controlled experiments with observational studies. which may point to causation: if exposure causes disease. and nonrandomized controlled experiments. Are the groups comparable? different? Were there confounding factors? adjustments? etc. .

Supplements The following slide provides supplements to our study of experiments and observational studies. .

will he grow taller by 20 × 0. Philippines.047 inches per lb § 0.9 inches? How does this case differ from a study of Hooke·s Law? (next slide) . using linear regression If a man puts on 20 pounds. study of IR8 The next two go ahead«  Predicting a student·s 1st-year GPA from her Math SAT. again using linear regression  Predicting a man·s height from his weight.Design is central in applied statistics« Which studies include an intervention? self selection? Which are observational? experimental?  Salk Vaccine Field Trial  Nurses Health Study of the ´effectsµ of long term oral contraceptive use  Women Health Initiative of HRT  IRRI.

and the wire lengthens by about 2 × 0.05 cm per kg) × load + 439. 1653-1703) Hooke·s Law« Hang a weight on a length of piano wire and it will stretch.Robert Hooke (England. it turned out that length § (0.05 = 0. . In one experiment at Berkeley.1 cm Increase the load by 2 kilograms.10 centimeter.