Case 5: Public Health (Case Control Studies)

Case Control Studies


Outline the characteristics of Case Control Studies


  • Case control is an Observational Study because there investigator does not control the exposure


  • Case control studies are Analytic studies because Case control studies contain a comparison group


Differentiating between "Cases" and "Controls"


Differentiate between "Cases" and "Controls"


  1. Cases refers to people with health outcomes/disease of interest


  1. Controls refers to people without the health outcome/disease of interest


  • In controls, in respect to the exposure and outcome of interest, they must resemble the cases that we are studying


  • So that the controls are people that are likely to have been cases if they had been exposed to the exposure of interest in the study


  • NOTE: This does not mean that controls cannot have a disease


  • It just that controls can NOT have the disease of interest
  1. Cross-sectional study
  • Cross-sectional study provides a snapshot of the exposure and the outcome at a fixed point (fixed period in time)


  • Advantages: Cross-sectional study is a fairly quick and relatively affordable way of describing the burden of disease


  • Disadvantages: Cross-sectional study cannot form a temporal relationship between the exposure and the outcome



    • Cannot say if the exposure came before the outcome


  1. Cohort Study
  • Cohort study provides a prospective view to determine if a disease or an outcome develops


  • Cohort study begins with a disease free population. This population may have other disease but it can NOT have the disease of interest


  • The selected population is then divided into an Exposed and an Unexposed group


  • Then all participants are followed-up over time to measure the development of an outcome in the 2 groups


  • Advantages:



    • Cohort study provides a temporal relationship between the Exposure and the Outcome


    • Allows for the measurement of risk of developing a disease based on specific exposure




  • Disadvantages:



    • Cohort Study is complex
    • It often requires long follow-up times
    • It is expensive
    • It is prone to high losses to follow-up

Case Control Study


Outline the Case Control Study


  • Case Control Study requires a retrospective view or running a backwards analysis from Outcome to Exposure


  • In a Case Control Study, from a population of interest, Cases and Controls are selected.


  • A history taking is done for each group in order to ascertain each person's exposure status


  • This results in 4 groups that are divided as follows:
  1. Cases with the exposure of interest


  1. Cases without the exposure of interest


  1. Controls who are exposed


  1. Controls who are not exposed

Key Features of Case Control Studies


Outline the key features of Case Control Studies


  • In Case Control Studies:
  1. Participants are selected according to their disease/outcome


  1. Researcher begins with the "cases", individuals who have the outcome of interest


  1. Researcher then selects an appropriate "control group", individuals who do not have the outcome of interest


  1. Once the "cases" and "controls" have been selected the researcher will have to collect information on the exposure(s) of interest in both groups


  1. Researcher then compares the results from the exposure history between the Cases and Controls


  1. Then the Researcher will then calculate a Measure of Association

Example of a Case Control Study


Research question: Is there an association between Smoking and Lung Cancer ?


  • Outcome: Lung Cancer


  • Exposure: Smoking




  • Process of Case-Control Study
  1. An specific population is identified from which Cases and Control Group are selected. For example a Hospital


  1. Cases are then selected, who are the Group of people who have Lung cancer


  1. Control group is then selected, who are the Group of people who do not have Lung cancer


  1. Researcher will then undertake a history-taking to look back in time and determine each person's smoking exposure


  1. From the results, the frequency of smoking is compared between the Cases and the Controls


  1. This will allow for the calculation of the Measure of Association

Considerations in the design of Case-Control: Selection of Cases


Outline the procedure involved in the Selection of Cases


  1. To select Cases, the study must make sure that they use a Standard Case Definition


  1. If the Case Definition is not Standardised, the may accidently include Non-Cases in the Case group


  • This can lead to invalid results


  1. Case definition may include Clinical symptoms, Laboratory Results, X-Ray findings etc


  1. Standardizing the Case definition means that you determine who will be included in the Case definition before the study begins


  1. Study should preferably use Incident (new) cases rather than Prevalent (old) cases


  • This is because the Diagnostic Criteria change over time


  • People who have had the disease for a long time: their exposure to certain risk factors may have changed over time as a result of the disease


  1. Sources of study population include: Patient Hospitals, Clinics, Disease Registries, Mass Screening and Cohort Studies

Considerations in the Design of the Case-Control: Selection of Controls


Outline the procedure involved in the Selection of Controls


  1. Control group needs to be free of the disease of interest


  1. Control group needs to be comparable to the Cases, meaning that they need to come from the same general/source population as the Cases


  1. Controls should represent people who would have been selected as Cases IF they had the disease of interest



    • This is the population at risk of becoming Cases


    • Usually similar in age, region or workplace to Cases




  1. Selection of Controls needs be independent of the exposure being investigated


  1. Sources of Controls include Hospitals, Clinics and Neighbourhood


  1. If the Controls are not selected properly, this will result in Selection Bias

Consideration in the Design of the Case-Control: Selection of Controls


Example of Bias introduced by Inappropriate Controls


  • Research question : Do NSAIDs prevent colo-rectal cancer?

Case- control study


  • Cases are selected from hospital oncology department


  • Controls are selected from rheumatology clinic based at hospital




  • All the participants are from the same hospital BUT different departments within the hospital


  • Could this choice of controls introduce any bias?


  • Possible Bias:
  1. Individuals at rheum clinic are more likely to be using NSAIDS that an individuals from the general population from which the cases came


  1. Exposure will be higher than that in the general population
  • This Selection bias will result in a reduced Association between the Exposure and the Outcome

Considerations in the Design of the Case-Control Studies: Determining the Exposure


Outline the procedure involved in determining the Exposure


  1. Exposure data can be collected at Face-to-Face Interviews, Telephone, Postal and Laboratory Tests


  1. Methods used must be the SAME for Cases and Controls


  1. Misclassification of Exposure will introduce bias


Outline some of the difficulties faced when determining the Exposure in Case-Control Studies


  1. Recall Bias
  • There is a potential for difference in recall between Cases and Controls


  • It is then important to try to reduce this by using Memory aids such as Photographs, Calendars with Major events


  1. Bias in Methods used to Collect Data
  • There is a potential for data collectors to determine exposure status differently in Cases and Controls


  • If possible data collectors should be blinded to disease status, to hypothesis and trained to elicit exposure information in a standard and structured way

Measuring the association between Disease and Exposure


Outline how to measure the association between Disease and Exposure


  • We cannot directly estimate the Incidence of Disease in the Exposed and Unexposed


  • We can measure the Odds of Exposure among the Cases and Controls


  • Odds = Number Events/Number of Non-Events


  • Measure of Association in Case-Control Studies is the Odds Ratio (OR)


  • Odds Ratio = Odds of exposure among Cases/ Odds of Exposure among Controls


  • Interpretation of Odds Ratio (OR) is similar to that of Relative Risk (Risk Ratio)


Calculation of the Odds Ratio


Outline the Calculation of Odds Ratio


  • Odds Ratio (OR) is a ratio in which the numerator is not included in the denominator


  1. D represents the Case and Controls



    • D+ represents Cases


    • D- represents Controls




  1. E represents Exposure



    • E+ represents Having the Exposure of interest


    • E- represents Not having the Exposure of interest




  1. calculation of Odds of Exposure

  • Odds of Exposure among Cases = Number of Cases exposed/ Number of cases not exposed



    • Odds of Exposure among Cases = a/c


  • Odds of Exposure among Controls = Number of Controls exposed/ Number of Controls unexposed



    • Odds of Exposure among Controls = b/d


  1. Odds Ratio

= Odds of exposure in Cases/ Odds of exposure in Controls


= (a/c)/ (b/d)


= (a/c) x (d/b)


= (ad)/ (cb)

Interpreting the Odds Ratio


Explain the interpretation of the Odds Ratio


  1. If the Odds Ratio (OR) = 1, then the Odds of Exposure is same in both the Cases and Controls


  1. If the Odds Ratio (OR) > 1, then the Cases have Greater Odds of being exposed than the Controls


  • The exposure is likely to be a risk factor


  1. If the Odds Ratio (OR) < 1, then the Cases have Lower Odds of being exposed than the Controls


  • The exposure is likely to be a Protective Factor


For Example: Interpret the following Odds ratio


  1. OR = 2.5


  • The Odds Ratio (OR) is greater than 1, then the Cases are 2.5 times more likely than the Controls to have been exposed


  • Exposure is likely a risk factor


  1. OR = O.5
  • The Odds Ratio (OR) is less then 1, then the Cases are half as likely to have been exposed compared to the Controls


  • Exposure is likely a protective factor

Example calculate the Odds Ratio


Exposure: Smoking


Outcome: Lung cancer


  • a = 127
  • b = 35
  • c = 27
  • d = 165


  1. Odds of exposure among the Cases
    = (a/c)
    = (127/73)
    = 1.74


  1. Odds of exposure among the Controls
    = (b/d)
    = (35/165)
    = 0.21


  1. Odds Ratio
    = (a/c)/ (b/d)
    = 8.20


  • Odds of exposure among cases was 8 times the odds of exposure among controls


  1. Therefore, the Odds ratio is greater then 1, The Lung cancer patients are 8.20 times more likely to have been exposed to smoking compared to the Non-Lung Cancer Patients


  • There is an association between the Smoking and Lung Cancer. Smoking is likely to be a risk factor for Lung Cancer



    image

Example calculate the Odds Ratio


  • Exposure: Diet high in vegetables


  • Outcomes: Obesity




  • a = 121
  • b = 171
  • c = 129
  • d = 79


  1. Odds of exposure among the Cases

= (a/c)
= (121/129)
= 0.938



  1. Odds of exposure among the Controls

= (b/d)
= (171/79)
= 2.16



  1. Odds Ratio

= Odds of Exposure among the Cases/ Odds of Exposure in Controls
= (121/79)/ (171/79)
= 0.43


  • Odds of exposure among the Cases was 0.43 times the Odds of Exposure among the Controls


  1. The Odds Ratio is less than 1, the individuals with Obesity are 0.43 times more to have a diet high in vegetables than the individuals that are not obese.


  • A diet high in vegetables is most likely a protective factor against obesity



    image

Precision


Outline the concept of Precision in Case-Control Studies


  • Odds Ratio is calculated from a Sample of the Population and no the entire Population


  • Therefore, in order to get the Precision on how well the Odds Ratio estimates the True Odds in the population, we need to interpret the Confidence Intervals (CI)


Explain the interpretation of Confidence Intervals (CI) in the Case-Control Study


  • Confidence Intervals (CI) tell us how confident we are that the results we have obtained are not due to chance


  • But represent the True Odds in the population


  • 95% means that we are 95% confident that the Estimate (True Odds Ratio) falls within the specific range


  • For Example: OR = 8.3 (95% CI 6.0 – 9.5) means that


  1. Case are 8.3 times more likely to have been exposed compared to the Controls


  2. We are 95% confident that the True Odds Ratio lies between 6.0 and 9.5




For example: OR = 0.43 (95% CI 0.65 – 0.25) means that


  1. Cases are 0.43 times more likely to have been exposed compared to Controls


  1. We are 95% confident that the Tur Odds Ratio lies between 0.65 and 0.25

Confidence Intervals



  • If we were to repeat the study a 100 times and calculated a 95% confidence interval each time, the true estimate would lie in the Confidence Interval 95/100 times


  • The purpose of the is Confidence Interval is that the Confidence Intervals helps interpretation due to uncertainty that results from random variation due to sampling

Example of a Case-Control Study


  1. Case-control study to determine risk factors for Salmonella Enteridis
  • Salmonella lives in intestinal tracts of humans and other animals and birds. Usually transmitted to humans by eating food contaminated with animal faeces


  • Causes gastrointestinal illness characterized by fever, sudden onset of headache, abdominal pain, diarrhoea and sometimes vomiting.


  • Typically, illness last for 5 to 7 days – but sometimes can cause severe dehydration and death unless treated promptly with antibiotics


  • Incubation period 12 – 36 hours – up to 1 week.


  1. Cases – patients with laboratory confirmed S Enteridis infection. They were enrolled shortly after the diagnosis.


  • Controls - persons who lived in the same neighbourhood as the cases and were a similar age with no diarrhoeal illness in the past 4 weeks. 2 controls were enrolled for each case


  • Standard questionnaire used to collect data on foods and beverages consumed, recent travel and recent food handling practices. Face to face interview.


image



  1. Which exposures appear to be risk factors for S Enteridis?
  • Eating dishes that contained Raw or Undercooked Eggs


  • This is because the Odds Ratio is greater than 1, and the individuals with Salmonella Enteridis are 18.9 to have eaten dishes that contained Raw or Undercooked Eggs compared to those who do not have S Enteridis


  1. Which exposure are protective ?
  • Eating chicken
  • Buying refrigerated eggs
  • Refrigerating eggs at home


  • Eating ground beef, Exposed to live chickens


  • This is because the CI include 1 and therefore the findings are not significant

Case-Control Studies: Strengths


List the Strengths of Case-Control Studies


  • Case-Control Studies:
  1. Relatively quick and inexpensive compared to Analytic Study Designs


  1. Well-suited in the evaluation of disease with Long-Patent Periods
  • These are diseases where the time period between the Exposure and Disease development is extended such as HIV


  1. Can assess multiple risk factors for a single disease


  1. Efficient for studying Rare disease (Compared to Cohort Studies).


Case-Control Studies: Limitation


List the Limitations of Case-Control Studies


  • Case-Control Studies:
  1. Inefficient for evaluating rare exposures


  1. Particularly prone to bias such as Selection Bias, Recall Bias, and Measurement Bias


  1. Cannot directly calculate Incidence Rates (compared to Cohort Studies and RCTs)


  1. More difficult to establish temporal relationship (Compared to Cohort Studies and RCTs)
  • Eg: Exposure occurred before outcome

Cofounding


Explain the concept of Cofounding


  • Cofounding is a type of bias which occurs in Observational Studies when a 3rd Variable (other than the Outcome and Exposure) results in an apparent association between the Exposure and Outcome


  • A cofounder can either increase or decrease the association between the Exposure and Outcome


An Example of Cofounding


  • Study looks at whether Coffee drinking causes Lung Cancer


  • After studying the results that confirm the Association between Drinking Coffee and the development of Lung cancer disease


  • A 3rd Variable (Cofounding Variable) is considered, where this variable has an association between the Exposure(Coffee drinking) and the Outcome (Lung cancer)


  • Majority of Cigarette smokers drink coffee and Cigarette smoking causes Lung Cancer


  • Therefore, the association between Coffee drinking and Lung cancer looks high because there is also an association between Coffee drinking and smoking

Cofounding


Outline the features of a Cofounder Variable


  • For a variable to be a Cofounder it must:
  1. be an independent determinant of the Disease/Outcome


  1. be associated with the exposure under investigation


  • Cofounding can be controlled either in the design of the study or in the analysis

Study Design 1


A study was conducted to investigate the association between having asthma and owning a pet. The hypothesis is that owning a pet is protective. Participants were randomly selected from a phone book by residential address and asked whether they had asthma and owned a pet.



  • What is the exposure ?
  • What is the outcome ?
  • What study design is being used?
  • What measure of association will be generated?
  • To show a protective effect should the measure of association be < 1 or > 1


Study Design 2


A study was conducted to examine the relationship renal cancer and dry-cleaning solvents. Patients with renal cancer were selected from a cancer clinic and compared them with patients admitted to the orthopaedic ward with respect to exposure to dry-cleaning solvents.


  • What is the study design and why?
  • What measure of association will this study generate ?