Disadvantages of Case Control study

[MudPhud20XX]

Hi all, so I always get confused about defitions of these study types and their limits.
FC lists both advantages/disadvantages of the Case Control study as below. I put my questions in bold and would appreciate if anyone can help me out. In general, I think I lack understanding of these studies and if anyone can help me out I would really appreciate it.

• Advantages to case control studies include:
  • high statistical power for rare diseases --> Why? Compared to Cohort study?
  • samples are easy to collect in a clinical setting -->Why? b/c it's retrospective and you just need to reveiw charts compared to Cohort study?
  • cost effective
• Disadvantages to case control studies include:
  • the possibility of selection bias --> Again is this b/c it's a retrospective study?
  • the possibility of confounding factors --> not really sure why this would be particularly a problem in Case control
  • results of the study do not provide a direct estimate of risk for developing a disease --> Again not sure why

---

Members don't see this ad.

 

[seminoma]

FWIW I didn't know any of those things and biostats/epi was one of my strongest areas. Don't get caught up in this much detail unless it's for your own personal enjoyment.

 

[Phloston]

Hi all, so I always get confused about defitions of these study types and their limits.
FC lists both advantages/disadvantages of the Case Control study as below. I put my questions in bold and would appreciate if anyone can help me out. In general, I think I lack understanding of these studies and if anyone can help me out I would really appreciate it.

• Advantages to case control studies include:
  • high statistical power for rare diseases --> Why? Compared to Cohort study?
  • samples are easy to collect in a clinical setting -->Why? b/c it's retrospective and you just need to reveiw charts compared to Cohort study?
  • cost effective
• Disadvantages to case control studies include:
  • the possibility of selection bias --> Again is this b/c it's a retrospective study?
  • the possibility of confounding factors --> not really sure why this would be particularly a problem in Case control
  • results of the study do not provide a direct estimate of risk for developing a disease --> Again not sure why

If a disease is ridiculously rare, it's virtually impossible to feasibly put together a cohort study comparing relative risk based on the presence/absence of a risk factor because you have no idea who is going to develop that (e.g.) 1/ten million disease. The most feasible way to do it is to seek out the people who already have that rare disease and look back in time at what may have contributed to its development. So case-control studies are good for rare diseases because they're most practical (because they save a ton of money and time). And looking back in time (a retrospective study) costs less because following people forward in time understandably, in and of itself, would have to require money and time.

There exists selection bias if a study is non-random in terms of whom you're selecting to be a participant in it. If you're selecting out people so you can evaluate them retrospectively, then of course that's not random. Any non-randomization weakens a study because of confounding. If you randomly select participants from the population (i.e., no selection bias), the confounding variables are also randomly distributed among your study participants. But any time you pick and choose who is in your study, there exists the chance you're inadvertently picking and choosing people who share a second variable. Because after all, we're doing research on a topic because we acknowledge we're not fully informed, so in turn we accept that we can't possibly be aware of all potential confounding.

A direct estimate of risk is hard to determine to begin with. Since you're looking back in time with a case-control study, not only are we non-random, but we also know people are poor historians. The inherent structure of the study acknowledges that we couldn't possibly be fully aware of other confounding/risk factors that may have popped up over time. Regardless, case-control studies determine odds ratio, which is AD/BC. Cohort studies determine relative (A/A+B / C/C+D) risk, which in turn can be used to calculate attributable (A/A+B - C/C+D) risk.

Selection bias on Step 1 is also notably due to when a researcher doesn't take into account loss to follow up.

 

[Keto]

Thanks for the good explanation, Phloston.
As Phloston mentioned, if you want to investigate a correlation between X and a rare disease Y (glioblastoma f.e.), you would need to follow millions and millions of people over a long period of time in order to get some power. This is simply not possible. Furthermore, a big difference between a cohort and a case-control study is, how the patients are selected: In case-control, patients are selected by disease (outcome). In cohort, patients are select by exposure (or risk factor) and then observed, until disease (outcome) occurs.
Retrospective cohort studies exist. Patients are again selected by exposure (or risk factor) that occurred in the past (f.e. patients that already smoked for 10 years are included in the cohort study and followed until disease (outcome) occurs).
One last small point for an exam. Try this mnemonic, if everything else fails: Cohort gives you relative risk. Case-control gives you odds ratio (calculated as mentioned above by Phloston).


Try these practice questions:

Suicides in teenagers in a small town had been a rare event before 20 cases were recorded in 2001. An investigation was started to try and determine the reason for this increase. The health care department suspects that the suicides are linked to increasing numbers of new families who have recently moved to town. The best type of study to investigate this would likely be a

1) Case-control study
2) Cohort study
3) Community trial study
4) Cross over study
5) Cross sectional study

A case-control study was designed to investigate the relationship between cigarette smoking and testicular cancer. The researcher interviewed patients with testicular cancer. They then selected the neighbors of the patients with the same age and race, and used them as controls. Such study design helps to control which of the following problems?

1) Ascertainment bias
2) Confounding
3) Observation bias
4) Recall bias
5) Selection bias

A case-control study was performed to investigate the relationship between colorectal cancer and family history of colorectal cancer. 80 cases with colorectal cases and 320 controls with no colorectal cancer were included into the study. Of the 80 cases, 60 had a positive family history of colorectal cancer. Of the 320 controls, 120 had a positive family history of colorectal cancer. Perform an adequate data analysis and calculate the estimated risk for patients having colorectal cancer with a positive family history.

1) 0.24
2) 3.6
3) 5
4) 24
5) 30

 

[MudPhud20XX]

Thanks for the good explanation, Phloston.
As Phloston mentioned, if you want to investigate a correlation between X and a rare disease Y (glioblastoma f.e.), you would need to follow millions and millions of people over a long period of time in order to get some power. This is simply not possible. Furthermore, a big difference between a cohort and a case-control study is, how the patients are selected: In case-control, patients are selected by disease (outcome). In cohort, patients are select by exposure (or risk factor) and then observed, until disease (outcome) occurs.
Retrospective cohort studies exist. Patients are again selected by exposure (or risk factor) that occurred in the past (f.e. patients that already smoked for 10 years are included in the cohort study and followed until disease (outcome) occurs).
One last small point for an exam. Try this mnemonic, if everything else fails: Cohort gives you relative risk. Case-control gives you odds ratio (calculated as mentioned above by Phloston).


Try these practice questions:

Suicides in teenagers in a small town had been a rare event before 20 cases were recorded in 2001. An investigation was started to try and determine the reason for this increase. The health care department suspects that the suicides are linked to increasing numbers of new families who have recently moved to town. The best type of study to investigate this would likely be a

1) Case-control study
2) Cohort study
3) Community trial study
4) Cross over study
5) Cross sectional study
So in this case, we would do cohort study? or should we do case-control? I feel like either can be done to see if suicide is related to new family members moving to town.

A case-control study was designed to investigate the relationship between cigarette smoking and testicular cancer. The researcher interviewed patients with testicular cancer. They then selected the neighbors of the patients with the same age and race, and used them as controls. Such study design helps to control which of the following problems?

1) Ascertainment bias
2) Confounding
3) Observation bias
4) Recall bias
5) Selection bias
I will cross out 2 and 4. Not sure what 1 is. I am narrowing my answer choice btw 3 and 5. The answer is 5 selection bias?

A case-control study was performed to investigate the relationship between colorectal cancer and family history of colorectal cancer. 80 cases with colorectal cases and 320 controls with no colorectal cancer were included into the study. Of the 80 cases, 60 had a positive family history of colorectal cancer. Of the 320 controls, 120 had a positive family history of colorectal cancer. Perform an adequate data analysis and calculate the estimated risk for patients having colorectal cancer with a positive family history.

1) 0.24
2) 3.6
3) 5
4) 24
5) 30
#3. 5, since this is case control, I made a 2X2 table and calculated the odd ratio.

 

[Keto]

Suicides in teenagers in a small town had been a rare event before 20 cases were recorded in 2001. An investigation was started to try and determine the reason for this increase. The health care department suspects that the suicides are linked to increasing numbers of new families who have recently moved to town. The best type of study to investigate this would likely be a

1) Case-control study
2) Cohort study
3) Community trial study
4) Cross over study
5) Cross sectional study
The answer would be case-control study. You take the cases (families with a case of suicide) and control (no suicide) and examine, whether the odds of the cases are higher in families with suicide cases. You want to have an answer/qlue fast and this is best done with a case-control study. A cohort study would not be the right answer.
Here are some examples of similar studies:
"Suicide in custody: case-control study." http://www.ncbi.nlm.nih.gov/pubmed/15572740
or "Firearms and adolescent suicide. A community case-control study." http://www.ncbi.nlm.nih.gov/pubmed/8213677
As a side note: Emile Durkheim, the "architect of modern social sciency", conducted a case study in 1897 by comparing suicide rates in different populations (https://en.wikipedia.org/wiki/Suicide_(book)).

A case-control study was designed to investigate the relationship between cigarette smoking and testicular cancer. The researcher interviewed patients with testicular cancer. They then selected the neighbors of the patients with the same age and race, and used them as controls. Such study design helps to control which of the following problems?

1) Ascertainment bias
2) Confounding
3) Observation bias
4) Recall bias
5) Selection bias
The answer would be confounding. The stem describes a matching process, which is frequently done for a case-control study. Matching is an efficient method to control confounding. Initially, matching variables are selected, which should always be the potential confounders (age, race, sex). Cases and controls are then selected based on the matching variables, such that both groups have a similar distribution in accordance with the variables. Selection bias is not controlled by matching, because the controls selected may not reflect the exposure experience (in this case, alcohol consumption) of the general population.
Here is a good summary, that explains, how confounding can be controlled by matching, randomization, and restriction: http://sphweb.bumc.bu.edu/otlt/MPH-...nfounding-EM/BS704-EP713_Confounding-EM5.html

A case-control study was performed to investigate the relationship between colorectal cancer and family history of colorectal cancer. 80 cases with colorectal cases and 320 controls with no colorectal cancer were included into the study. Of the 80 cases, 60 had a positive family history of colorectal cancer. Of the 320 controls, 120 had a positive family history of colorectal cancer. Perform an adequate data analysis and calculate the estimated risk for patients having colorectal cancer with a positive family history.

1) 0.24
2) 3.6
3) 5
4) 24
5) 30
5 is correct!

 

[MudPhud20XX]

Suicides in teenagers in a small town had been a rare event before 20 cases were recorded in 2001. An investigation was started to try and determine the reason for this increase. The health care department suspects that the suicides are linked to increasing numbers of new families who have recently moved to town. The best type of study to investigate this would likely be a

1) Case-control study
2) Cohort study
3) Community trial study
4) Cross over study
5) Cross sectional study
The answer would be case-control study. You take the cases (families with a case of suicide) and control (no suicide) and examine, whether the odds of the cases are higher in families with suicide cases. You want to have an answer/qlue fast and this is best done with a case-control study. A cohort study would not be the right answer.
Here are some examples of similar studies:
"Suicide in custody: case-control study." http://www.ncbi.nlm.nih.gov/pubmed/15572740
or "Firearms and adolescent suicide. A community case-control study." http://www.ncbi.nlm.nih.gov/pubmed/8213677
As a side note: Emile Durkheim, the "architect of modern social sciency", conducted a case study in 1897 by comparing suicide rates in different populations (https://en.wikipedia.org/wiki/Suicide_(book)).

A case-control study was designed to investigate the relationship between cigarette smoking and testicular cancer. The researcher interviewed patients with testicular cancer. They then selected the neighbors of the patients with the same age and race, and used them as controls. Such study design helps to control which of the following problems?

1) Ascertainment bias
2) Confounding
3) Observation bias
4) Recall bias
5) Selection bias
The answer would be confounding. The stem describes a matching process, which is frequently done for a case-control study. Matching is an efficient method to control confounding. Initially, matching variables are selected, which should always be the potential confounders (age, race, sex). Cases and controls are then selected based on the matching variables, such that both groups have a similar distribution in accordance with the variables. Selection bias is not controlled by matching, because the controls selected may not reflect the exposure experience (in this case, alcohol consumption) of the general population.
Here is a good summary, that explains, how confounding can be controlled by matching, randomization, and restriction: http://sphweb.bumc.bu.edu/otlt/MPH-...nfounding-EM/BS704-EP713_Confounding-EM5.html

A case-control study was performed to investigate the relationship between colorectal cancer and family history of colorectal cancer. 80 cases with colorectal cases and 320 controls with no colorectal cancer were included into the study. Of the 80 cases, 60 had a positive family history of colorectal cancer. Of the 320 controls, 120 had a positive family history of colorectal cancer. Perform an adequate data analysis and calculate the estimated risk for patients having colorectal cancer with a positive family history.

1) 0.24
2) 3.6
3) 5
4) 24
5) 30
5 is correct!

Keto, thanks so much for the explanation. With regard to #2, here is what FA says about confounding bias.

"When a factor is related to both the exposure and outcome, but not on the causal pathway Ž factor distorts or confuses effect of exposure on outcome."

Solely based on the question stem, I wasn't sure if this would qualify as confounding bias. How does investigating the relationship btw cigarette smoking and testicular cancer distort the confuses the outcome in this particular example of yours?

 

[Keto]

Hi...

so in this question they describe the process of matching, which is done in any case-control study and is an efficient way to control confounding. Why is that so? If you start with "matching", the first step would always be to define matching variables, which could be potential confounders. Cases and controls are then selected based on the matching variables, such that both groups have a similar distribution in accordance with the variables.

So if you select f.e. a neighbor as a "control", who has the same socioeconomic status, who is the same race, age and sex as the "case".. then you would potentially minimize confounding.

I hope this helps.