PS 1-4
All New York City rescue workers who took place in World Tower rescue operations on 9/11/2001 and who are being followed for respiratory disease. What is the population type? What is the most appropriate summary statistic to describe study outcomes?
Open population and incidence rate.
When is it desirable to use a case-control study?
a. When the disease is rare.
b. When the study population is dynamic.
c. When the disease has a long latency period.
An epidemiologist wants to study the efficacy of iron supplementation for the prevention of HIV infection. He wants to make sure that only subjects who are (still) free of HIV infection are enrolled in his trial. Therefore, he screens a large group of people using a diagnostic test. Based on the outcome of the test, he decides who can participate in his iron supplementation trial. For this purpose, it is very important that the diagnostic test has a high…
Negative Predictive Value
In a matched case-control study, do you need to control for matched characteristics in the statistical analysis?
Yes
When effect measure modification exists, is it meaningful to report a summary measure of the exposure-outcome association that pools data across different strata of the effect modifier?
NO!
Name 4 types of study that could potentially assess the temporal association between an exposure and outcome?
Retrospective, Prospective, Case-control, and RCT.
Assuming the size of the population at risk is declining over time, would CI = IR*T overestimate or underestimate the cumulative incidence? What method would be better for calculating the cumulative incidence?
CI = IR*T overestimates the cumulative incidence because there is an exponential decay in the population at risk. A more accurate mathematical expression that takes this into account is: CI = 1-e^(-IR*T).
A psychiatrist devised a short screening test for depression. An independent blind comparison was made with a gold standard for diagnosis of depression among 200 psychiatric outpatients. Among the 50 outpatients found to be depressed according to the gold standard, 35 patients were positive for the test. Among 150 patients found not to be depressed according to the gold standard, 30 patients were found to be positive for the test. Calculate the sensitivity, specificity, PPV, NPV, and prevalence.
Sensitivity = the proportion of subjects with the disease correctly diagnosed by the test = 35/50 = 70%.
Specificity = the proportion of subjects without the disease correctly excluded by the test = 120/150 = 80%.
Positive predictive value = the proportion of subjects with a positive test who have the disease = 35/65 = 53.8%.
Negative predictive value = the proportion of subjects with a negative test who do not have the disease = 120/135 = 88.9%.
Prevalence of depression = 50/200 = 25%
What three criteria are necessary to consider a covariate to be a potential confounder?
1) It is associated with the exposure of interest in the source population/among controls
2) It is associated with the outcome of interest among the unexposed
3) It is not on the causal pathway from the exposure to the outcome.
When lifelong smokers are told they have lung cancer or emphysema, many may then quit smoking. This change of behavior after the disease develops can make it seem as if ex-smokers are actually more likely to die of emphysema or lung cancer than current smokers. What is this an example of?
Reverse causality; this occurs when the probability of the outcome is causally related to the exposure being studied
List 2 potential limitations for a prospective cohort study.
1. Not efficient for studying rare outcomes.
2. Associations observed in the study are subject to unmeasured confounding.
Suppose that the hypothesis test that was conducted for the measure of excess risk resulted in a P value of 0.08. What is the interpretation of a P value of 0.08?
There is an 8% chance of obtaining a value as extreme or more extreme than the observed value given that there is no association between exposure and disease.
A psychiatrist devised a short screening test for depression. An independent blind comparison was made with a gold standard for diagnosis of depression among 200 psychiatric outpatients. Among the 50 outpatients found to be depressed according to the gold standard, 35 patients were positive for the test. Among 150 patients found not to be depressed according to the gold standard, 30 patients were found to be positive for the test. A general practitioner decides to use the screening test for his patients. The prevalence of depression among his patients is estimated to be 5%. Which of the following measures are identical whether the test is applied to general practice patients or to the psychiatric outpatients?
Prevalence of depression
Sensitivity
Specificity
PPV
NPV
If V intake could have been measured more accurately, and thus confounding by V more completely controlled, would you expect the adjusted RR associated with high serum beta carotene levels to have been:
Less than 0.8
0.8
Greater than 0.8
Why?
Sensitivity and Specificity
The prevalence of depression in general practice (5%) is lower than among the psychiatric outpatients (25%). Sensitivity and specificity are characteristics of a diagnostic test and do not change with prevalence of the population. Positive predictive value increases with the prevalence, while negative predictive value decreases with the test.
On a scrap piece of paper, please draw a confounding diagram for the following scenario:
Smoking in mothers is known to be associated with a lower birth weight in infants and that association is being studied. Coffee consumption is a potential confounder because increased consumption is has been shown to also be associated with a lower birth weight and coffee drinking is more prevalent among smokers.
Smoking ----(-)----> Birth weight
\ /
(+) (-)
\ /
Increased coffee consumption
Name strategies for minimizing confounding in the design of a study (provide 3) and in the analysis of a study (provide 2).
Design: Restriction, Matching, Randomization
Analysis: Stratification, Multiple variable regression analysis
What do the P value and 95% confidence intervals specifically represent?
The P value represents the likelihood that the observed result or one more extreme is due to chance given the null hypothesis is true. The 95% confidence interval represents a range of values within which the true association is estimated to fall with 95% confidence.
What can you do to increase power to detect a statistically significant association between exposure and disease?
You could increase the number of participants that you include in the study.
You could use a more lenient Type I error threshold.
Using data obtained in a prospective cohort study, let us say you observe that the incidence of cancer among persons in the upper fourth of the distribution of serum beta carotene is 60% of that of persons in the lower three-fourths. In that study, the intake of a particular vegetable, V, was ascertained. However, the instrument used to ascertain V did so but in a cursory way. Because V contains nutrients other than beta carotene that might reduce cancer risk, you would like to adjust for intake of V when assessing the association between beta carotene levels and cancer. When you do this, the RR associated with being in the upper fourth of the serum beta carotene distribution is now 0.8 instead of 0.6.
If V intake could have been measured more accurately, and thus confounding by V more completely controlled, would you expect the adjusted RR associated with high serum beta carotene levels to have been:
Less than 0.8
0.8
Greater than 0.8
Why?
3. Incomplete adjustment for the confounding variable, intake of V, will result in an adjusted relative risk that is spuriously close to the crude relative risk. Therefore, since incomplete adjustment led to an increase in the apparent relative risk of cancer associated with high levels of serum beta carotene - from an unadjusted relative risk of 0.6 to 0.8 - more complete adjustment would be expected to lead to a relative risk that would be higher still.
In a case-control study of the association between a western diet and joint health as measured by the need for a non-sports related joint replacement, it was found that a western diet was associated with poor joint health. It was also observed that among the controls, obesity was associated with following a western diet. Additionally, among the participants not exposed to a western diet, the obese patients were more likely to require a non-sports related joint replacement.
Is obesity a confounder in this study?
No (mediator on the causal pathway, western diet → obesity → poor joint health)
Indicate whether effect modification, confounding, or both are present:
If the stratum-specific estimates differ from one another, and they are both less than the crude estimate or if they are both greater than the crude estimate
If the stratum-specific estimates differ from one another, and the crude estimate is between the two stratum-specific estimates
How did you determine this for part 2?
1) If the stratum-specific estimates differ from one another, and they are both less than the crude estimate or if they are both greater than the crude estimate, then there is both confounding and effect modification.
2) If the stratum-specific estimates differ from one another, and the crude estimate is between the two stratum-specific estimates, then you need to pool the stratum-specific estimates (with a Mantel-Haenszel equation) to determine whether the pooled estimate is more than 10% different from the crude estimate.
Note that in this situation you are only pooling the stratum-specific estimates in order to make a decision about whether confounding is present; you should not report the pooled estimate as an "adjusted" measure of association if there is effect modification
A case-control study was conducted among males to examine the association between smoking and bladder cancer. Bladder cancer is a rare disease. Given an OR of 1.67, estimate the percent of bladder cancer cases among male smokers that can be attributed to smoking and interpret this in words. Also, estimate the percentage of bladder cancer cases that could have been prevented in the population if smokers did not smoke, provided that the proportion of cases exposed is 0.86.
Use the relationship of AR% = (RR-1)/RR.
AR%= (1.67-1)/1.67= 0.40
Assuming a causal association between smoking and bladder cancer, 40% of bladder cancer cases among male smokers can be attributed to smoking.
PAR% = [(RR-1)/RR] * proportion of cases exposed
PAR% = 0.40*0.86= 34% of bladder cancer cases in the population can be attributed to smoking.
Calculate and interpret the vaccine efficacy for the following results:
The incidence of infection Q in the vaccinated group was 20/2000 and the incidence of infection with the disease in the unvaccinated group was 175/2000.
(Iunvacc. - Ivacc.)/ Ivacc.
= (175/2000 – 20/2000)/(175/2000)
= 88.57%
Of those unvaccinated, 88.6% of the infection with X is due to their lack of vaccination.
In a retrospective cohort study conducted in Denmark (Dalton et al., 2006), the incidence of cancer was monitored in persons who had been prescribed an anti-psychotic drug and compared to the incidence in other persons. While the investigators were unable to ascertain smoking histories on any of the cohort members, they did have information on hospitalization for chronic obstructive pulmonary disease (COPD, a condition related to smoking) that occurred among these persons during the follow-up period. The risk of lung cancer in users of anti-psychotic medications was 1.57 times that of non-users (95% confidence interval 1.40-1.75); after adjustment for hospitalization for COPD, the corresponding relative risk was 1.24 (95% confidence interval 1.10-1.40).
When examining a potential association between exposure and disease, it is generally not advisable to adjust for factors that occur after the exposure has begun. Nonetheless, you believe that the relative risk that better characterizes the relationship between use of an anti-psychotic drug and the incidence of lung cancer is the one that adjusts for hospitalization for COPD. Why? Had the investigators been able to adjust for smoking history as well as hospitalization for COPD, would you anticipate the relative risk to be larger, smaller, or the same as 1.24? Why?
The COPD-adjusted relative risk is the more valid one because: (1) It is likely that hospitalization for COPD is a result of smoking prior to (as well as after) the receipt of anti-psychotic medications; and (2) it is much less likely that hospitalization for COPD represents a consequence of use of an anti-psychotic medication (the only circumstance under which it would not be advisable to adjust for this variable).
Adjustment for hospitalization for COPD has almost certainly not entirely removed the confounding effect of smoking. The ability to adjust for smoking as well would almost certainly produce a still smaller estimate of the relative risk relating use of anti-psychotic drugs to the incidence of lung cancer.
What is used to determine the weights for the strata in a Mantel-Haenszel adjustment?
The inverse of the variance for each stratum
In the following scenarios, when is it appropriate to use Mantel-Haenszel methods to combine the stratum-specific measures of association into a single pooled measurement? Explain.
If there is only confounding?
If there is neither confounding nor effect modification?
If there is only effect modification?
If there is only confounding: The stratum-specific measures of association will be similar to one another, but they will be different from the overall crude estimate by 10% or more. In this situation, one can use Mantel-Haenszel methods to calculate a pooled estimate (RR or OR) and p-value.-- YES
If there is neither confounding nor effect modification: The crude estimate of association and the stratum-specific estimates will be similar. They don't have to be identical, just similar. --YES
If there is only effect modification: The stratum-specific estimates will differ from one another significantly. Effect modification is a biological phenomenon that should be described, so the stratum-specific estimates should be reported separately. -- NO