1.40 p=0.145

1.40p=0.145

Epidemiologic Association

• Chance

• Bias

• Confounding

• Truth

CHANCE

• “…the results of the retrospective cohort study was inconclusive with all but two of the data being non-significant.”

• “…analysis of self-reported cancer rates for children who had played at the site (Exhibit E) shows that while that was a positive association, the results were not statistically significant. “

1387 study participants, 73 self reported cancers

Self-Reported Cancer

Yes No

Nyanza Site Yes 36 547 236

Waste Ponds & Lagoons

No 36 767 1,151

72 1314 1386OR = 1.40 (95% CI 0.87, 2.25)

p=0.16

Assume MDPH Study ResultsWere as Below

What Does p=0.16 mean?

OR=1.40

P=0.08 P=0.0895% CI 0.87, 2.25

P=0.16


Yes No



No 36 767 1,151

72 1314 1386OR = 1.40 (95% CI 0.87, 2.25)

p=0.16


Yes No

Nyanza Site Yes 72 1,094 1,166


No 72 1,534 1,606

144 2628 2772

OR = 1.40 (95% CI 1.00, 1.96)p=0.047

Assume MDPH Study ResultsWere as Below

50% Participation

What if MDPH StudyHad Full Participation?

X 2

What is Bias?• Bias means that there is a systematic error in the

design or results• The measured (or estimated) variable is different

from the truth.• Depends on the question!

– Biased sample sample does not represent population– Biased exposure exposure ≠ true value– Diagnostic bias Disease ≠ true disease – Epidemiologic Bias association ≠ truth

• NOTE: Biased sample, or exposure, or disease does not mean association is biased (epidemiologic bias)!!

SELECTION BIAS

• “Selection bias may play a significant role if the 1,387 people out of 2,054 who agreed to participate were more likely to have cancer compared to those who declined participation.”

• “… individuals who may have already died either from cancer or unknown unrelated causes and thus were untraceable would significantly bias our findings.”

• “There is a chance of those people with cancer having enrolled in the study versus those who did not have cancer.”


Yes No



No 36 767 1,151

72 1314 1386OR = 1.40 (95% CI 0.87, 2.25)

p=0.16

OR = 1.40 (95% CI 1.00, 1.97)p=0.052

What if 50% of the Cancer Cases Died Before MDPH Study ?

Ca X 2Actual Cancer

Yes No



No 72 767 839

144 1314 1458

Assumed MDPH Study Results


Yes No



No 36 767 1,151

72 1314 1386OR = 1.40 (95% CI 0.87, 2.25)

p=0.16

OR = 1.40 (95% CI 0.93, 2.11)p=0.104

What if Exposed People Were Twice as Likely to Participate ?

Ca X 2



Yes No

Nyanza Site Yes 72 1,094 1,166


No 36 767 803

108 1,861 1969

SELECTION BIAS

• “There is a chance of those people with cancer having enrolled in the study versus those who did not have cancer.”

• Only 40 out of 73 cancers confirmed


Yes No



No 36 767 1,151

72 1314 1386OR = 1.40 (95% CI 0.87, 2.25)

p=0.16

OR = 1.39 (95% CI 0.74, 2.60)p=0.30

What if only 40 Confirmed Cancer Cases Included ?


Confirmed Cancer

Yes No



No 20 783 803

40 1346 1386

SELECTION BIAS

• Possible sources:– Selection of eligible participants– Differential participation– Loss to follow-up

• Requires:– Selection based on both exposure and

outcome

• SELECTION DOES NOT IMPLY SELECTION BIAS

INFORMATION BIAS

• “The difference between self-reporting and confirmed cases especially in view of the prolonged resentment held by the community members towards Nyanza lends some doubt to the results.”

CONFOUNDING BIAS

• “A number of potential confounders have been considered, others such as the possibility of other or previously existing sources of contamination have not been ruled out.”

• Age, sex, race ?

1.40p=0.145

1.17p=0.50

0.78p=0.58

1.35p=0.20

1.17p=0.50

1.36p=0.17

1.47p=0.106

1.26p=0.31

1.82p=0.176

Confounding Bias

• Confounding bias results when the association between exposure and disease is altered (biased) by a third variable (i.e. the confounder)

• Confounder by definition:– An independent determinant for the outcome (i.e. it is associated

with the outcome among the unexposed) AND – Associated with the exposure of interest through a noncausal

mechanism.

• Confounding– Inevitable in all studies. It is not investigator induced; it is a basic

characteristic of nature.– Can be minimized, but it is never possible to know effects of all

confounders– Best one can do is design study to reduce confounding and

collect information on known or suspected confounders to control in analysis.

Evaluating of Confounding

• Crude vs. Adjusted Associations– Confounding bias is the difference between the crude (unadjusted)

effect estimate and the effect estimate adjusted for the posited confounder.

– Rule of Thumb: at least a 10% change in effect estimate produced by adjustment for a true confounder

• Test of significance– Have no role in the identification or assessment of confounding

– The association of the confounder with the outcome or exposure does not have to be statistically significant to produce a substantial difference between crude and adjusted effect estimates.

• Residual Confounding– Confounding Bias remains after inclusion of the confounder in the

analysis due to mismeasurement of confounder or inadequate model for the confounder

Control of Confounding in Design

• Randomization– Random allocation of exposure (clinical trial)– Cannot (unlikely) to be realized in environmental epidemiology

• Restriction– Restrict participation by potential confounder

• No confounding if potential confounder does not vary across subjects

– Straight forward and inexpensive– Reduces number of subjects (and power)– Residual confounding possible– Limits generalizability (extrapolation)

• Improves validity of study

• Matching– Reduces confounding in cohort study– Induces confounding in case control

• Must control for matching variables

– Improves efficiency to control in analysis

EFFECT MODIFICATION

• “We already know that genetics plays a large role in many cancer etiologies. Family history is a significant confounder within our findings.”

Effect Modification

• Variation in the association between outcome and exposure across levels of some other variable.

• Confounders can also be effect modifiers.

Epidemiologic Association

• Chance

• Bias

• Confounding

• Truth

CAUSALITY

• “… the biological plausibility and the latency period of each type of cancer mentioned needs to be established in this population.”

• “Dose-response relationships need to be strengthened.”

CAUSALITY

• “We would recommend the use of GIS mapping techniques, a more detailed qualitative assessment of exposure histories among cases and controls in Ashland and the consideration of other communities with similar exposures to liquid hazardous waste during the same time period.”

• “Coordination with other communities with similar exposures across the country may serve to increase the numbers of exposures and cancers and improve statistical analysis.”

AUSTIN BRADFORD HILLThe Environment and Disease: Association or

Causation?Proc Royal Soc Med 1965; 58:295-300.

• Strength• Consistency• Specificity• Temporality• Biological

Gradient

• Plausibility• Coherence• Experiment• Analogy

Hill Criteria for Causation

• STRENGTH– First upon my list I would put the

strength of the association.– In thus putting emphasis upon the

strength of an association we must, nevertheless, look at the obverse of the coin. We must not be too ready to dismiss a cause-and-effect hypothesis merely on the grounds that the observed association appears to be slight.

Top


• CONSISTENCY.– Has it been repeatedly observed by

different persons, in different places, circumstances and times?

– ...there will be occasions when repetition is absent or impossible and yet we should not hesitate to draw conclusions.

Top


• SPECIFICITY– ...specificity of the association, the third

characteristic which invariably we must consider.

– ...if specificity exists we may be able to draw conclusions without hesitation; if it is not apparent, we are not thereby necessarily left sitting irresolutely on the fence.

Top


• TEMPORALITY– ...temporal relationship of the

association - which is the cart and which is the horse?

Top


• BIOLOGICAL GRADIENT– ...if the association is one which can

reveal a biological gradient, or dose-response curve, then we should look most carefully for such evidence.

Top


• COHERENCE– ...cause-and-effect interpretation of our

data should not seriously conflict with the generally known facts of the natural history and biology of the disease...

Top


• EXPERIMENT– ... Occasionally it is possible to appeal to

experimental, or semi-experimental, evidence. For example, because of an observed association some preventive action is taken.

Top


• PLAUSIBILITY– It will be helpful if the causation we

suspect is biologically plausible. But this is a feature I am convinced we cannot demand. What is biologically plausible depends upon the biological knowledge of the day.

Top


• ANALOGY– In some circumstances it would be fair

to judge by analogy.

Top


• TESTS OF SIGNIFICANCE– No formal tests of significance can

answer those questions. Such tests can, and should remind us of the effects that the play of chance can create, and they will instruct us in the likely magnitude of those effects. Beyond that they contribute nothing to the 'proof' of our hypothesis.


• CONCLUSIONS– All scientific work is incomplete -

whether it be observational or experimental. All scientific work is liable to be upset or modified by advancing knowledge. That does not confer upon us a freedom to ignore the knowledge we already have, or to postpone the actions that it appears to demand at a given time.

CAUSALITY

• “It is crucial that with further testing and investigations, we do not ignore action that must be taken in order to minimize exposure to the residents of Ashland.”

1.40 p=0.145

Documents

waste ponds lagoonsno36767

waste ponds lagoonsno20783

waste ponds lagoonsno72767

waste ponds lagoonsno721

mdph study resultswere

reported cancer rates

assumed mdph study resultsself

study participants