Impact of Depressive Symptoms on Hypertension Risk in Women

Study Purpose and Design

Review the following paragraph and respond to the questions below using a few sentences for each response. 

At baseline, 1,489 women were experiencing depressive symptoms and 4,821 women were not experiencing depressive symptoms based on the Beck Depression Inventory.  This questionnaire is a 21-question multiple-choice self-report questionnaire that was administered to all women in this study.  Participants were followed from 2010-2018 for incident hypertension.  Each year, participants completed a questionnaire to indicate whether they received a clinician-made diagnosis of high blood pressure in the previous year.  Participants were asked to write their systolic and diastolic blood pressure measurements and to answer other health-related questions. 

After adjustment for age, alcohol use, and body mass index, the following results were reported:

1.Depressive symptoms and risk of hypertension: RR (relative risk) = 1.42, 95% CI: 0.98-1.67

2.Depressive symptoms for >3 years and risk of hypertension: RR = 1.54, 95% CI: 1.21-1.89

3.The relationship between depressive symptoms and risk of hypertension was attenuated among women who began antidepressant treatment: RR=1.06, 95% CI: 0.73-1.37

a. What is the purpose of this study?  What is the exposure and what is the outcome? 

b. What is the study design?  On what basis did you determine the study design? 

c. What are the study findings?  What can you conclude based on these results?

d. Interpret the three different relative risks presented in the above paragraph within the context of the question.  Indicate whether any of the results are statistically significant and how you determined this.

e. Describe the potential for each of the following. (10 points) 

1.Nondifferential misclassification of exposure

2.Nondifferential misclassification of outcome

3.Confounding

4.Bias 

Indicate “true” or “false” for each of the questions below and explain why.

a.Issues of internal validity include generalizability. 

b.Confounding can only be addressed in the study design stage of the study. 

c.If interaction/effect modification occurs, it is best to control for this in the analysis stage. 

d.To avoid selection bias, controls in a case-control study must have been eligible to be cases should they have developed the disease.

e.A confounder is not on the causal pathway between the exposure and the outcome. 

f.A participant is a non-smoker at baseline.  Follow-up questionnaires did not ask about smoking status to capture that this participant became a smoker halfway through the study.  This would be an example of effect modification. 

3.  For each of the following questions, mark an “X” in the __ space for the statement that best answers the question.  For this section, choose one best answer for each question. 

Study Findings

a. Which of the following of Bradford Hill’s criteria is essential to proving a causal relationship?

__ Biological plausibility

__ Consistency/generalizability          

__ Specificity

__ Strength of association

__ Temporality (time sequence)

b. In a study of cell phone use and brain cancer, the questionnaire asks if participants have used a cell phone in the past year, and participants must answer “yes” or “no.”  Additional questions such as the duration of use or the type of cell phone used were not included on the questionnaire.  This could result in potential…

__ Nondifferential misclassification of exposure

__ Effect modification

__ Nondifferential misclassification of outcome       

__ Confounding

__ None of the above

c. In a case-control study of height and myocardial infarction, you observe that short men have two times greater risk of a myocardial infarction as compared to tall men.  In your study, you also observe that short men are older than tall men.  It is known that age is a risk factor for myocardial infarction.  In this study, age would be considered a…

__ Confounding variable

__ Matching variable 

__ Exposure variable

__ Source of bias

__ None of the above

d. In a cross-sectional study, you…

__ Start with exposed and non-exposed and follow participants for the outcome of interest

__ Start with diseased and non-diseased and follow participants for the exposure of interest

__ Start with exposure and disease information from participants at a single point in time

__ Randomize participants to different treatments and follow for the outcome of interest

__ None of the above

e. Which of the following is most applicable to a case-control study? 

__ The proper selection of an adequate control group is critical

__ Case-control studies are best to examine rare exposures

__ Case-control studies are less prone to bias compared to cohort studies   

__ The relative risk is the preferred measure of association

__ All of the above

4. For each of the scenarios below, indicate:

What type of bias or nondifferential misclassification occurred and why.  The options include: nondifferential misclassification of exposure, nondifferential misclassification of outcome, selection bias, recall bias, information/observation bias, or surveillance bias. 

The possible impact of this bias or nondifferential misclassification on the study results.  The options include: No influence, bias toward the null (underestimate), or bias away from the null (overestimate).

a. In a cohort study, men are asked at baseline to indicate their drinking history as: current, past, or never.  During follow-up, current drinkers are asked to meet with a study physician annually to check liver enzyme levels, the outcome of interest.  Past or never drinkers are asked to meet with a study physician once per three years to check liver enzyme levels. 

Interpretation of Relative Risks

Type of bias and/or nondifferential misclassification and why

Impact on study results 

b. A case-control study is conducted on the association between workload during pregnancy (categorized as: 40 hours/week) and the risk of preterm birth.  Women who worked >40 hours/week and experienced a preterm birth were more likely to participate in the study as they were concerned the preterm birth may have been caused by their excessive workload.

Type of bias and/or nondifferential misclassification and why

Impact on study results

5. For each of the following questions, choose all that apply by marking an “X” in the space for the statement(s) that answer each question.  For these questions, you may mark one or more answer for each question.

a. Bias in epidemiologic studies is best described as…

__ The effect of differential misclassification that cannot be corrected after it occurs

__ A problem with case-control studies that can be addressed with masking/blinding

__ Random variation that is present in all studies and does not influence study results

__ An influence of confounding where a third variable distorts the potential association

__ A critical flaw in an epidemiologic study that can be corrected after the analysis stage 

b. Strengths of a cohort study include…

__ Efficient for studying rare diseases

__ Randomization of participants to different treatments

__ Potential to study multiple outcomes

__ Establish a temporal relationship between the exposure and outcome

__ Relatively inexpensive

c. The association between alcohol intake and cancer risk has been a recent topic of interest.  Which of the following statements strengthen the association between alcohol and cancer risk and move the evidence toward the direction of a causal relationship?

__ The risk of cancer increases as the amount of alcohol exposure increases

__ Animal studies have shown evidence of cancerous lesions after alcohol exposure

__ Results from case-control studies of this exposure and outcome are not consistent

__ Cross-sectional studies have examined the temporality of this exposure and outcome

__ All of the above

d. Confounding can best be described as…

__ A major concern in randomized controlled trials based on masking/blinding

__ A mixture of effects from a third variable on the exposure-outcome relationship of interest

__ A bias that is only present in prospective cohort studies

__ A problem in epidemiologic studies that can be controlled for using statistical techniques

__ None of the above

e. Which of the following statement(s) relate to analytical epidemiology? 

__ Includes cohort and case-control study designs

__ Includes case report and ecological study designs

__ Generates hypotheses

__ Tests hypotheses

__ Focused on person, place, and time

6. A study recruited participants for a long-term follow-up study to examine the effects of exposure to a potentially dangerous pesticide.  At baseline, participants were categorized as exposed or non-exposed and disease status was determined throughout follow-up as participants attended annual visits with the study physician and completed a questionnaire. 

a. What is the study design?  On what basis did you determine the study design?

b. Calculate the relative risk of disease on being exposed to the pesticide at any level? Interpret this relative risk in the context of the question. 

c. Using the non-exposed participants as the reference group, calculate the relative risk of disease among participants with heavy exposure.  Interpret this relative risk in the context of the question.

d. Calculate the odds ratio of disease among participants with moderate exposure if the non-exposed participants serve as the reference group. Interpret this odds ratio in the context of the question.

e. You noticed that exposure status to this potentially dangerous pesticide was asked at baseline but not assessed in the follow-up questionnaires.  Discuss whether this could impact study results and how.