# Is my study case-control or retrospective cohort?

#### victorxstc

##### Pirate
Hi

AFAIK if we split the sample by the exposure, it is retrospective cohort; and if we split it by the disease, it is case control.

I am studying the association between malocclusion and the efficacy of mastication. I have divided the sample into two groups: people with and without malocclusion; I wanna check whether those with malocclusion have poorer mastication efficacy.

I think malocclusion is the exposure here, and then my study is retrospective cohort. However, my statistician says it is a case control, because we are dividing the sample according to the disease.

My main problem is that both the "malocclusion" and "mastication efficacy" are some kinds of diseases, although the mastication efficacy cannot be the cause of the malocclusion, but the malocclusion can be the cause of the other variable.

So I still think although the malocclusion variable itself is a disease, it is the exposure here and hence, my study is cohort.

Can somebody say am I right?

##### Ninja say what!?!
Hi

AFAIK if we split the sample by the exposure, it is retrospective cohort; and if we split it by the disease, it is case control.
It is more than just that. If you have a cohort that you are following, then it's a cohort. You can split it by exposure once you know it's a cohort. If you are following it forward in time, then it's prospective. Otherwise, it's retrospective.

If you only have a whole bunch of cases of diseases and you're trying to match the same amount of controls, then it's case control. Here, you're still splitting them by exposure. The difference is that you have a set of cases first and you try to match controls after.

It looks like you have a group of people who were collected from the same place. All you're doing is dividing them by whether they were exposed or not and seeing what the disease rate is in each group. If this is the case, then I would say also that it is a retrospective cohort.

In order for me to be sure though, I'd like to know more about the data you have.

#### victorxstc

##### Pirate
Wow, Thank you so very much for your prompt help.

About the retrospective cohort vs. prospective one, thank you very much.

About the study protocol:

Malocclusion means when someone has some problems in his teeth well occluding together. It is a disorder.
Mastication performance means how effectively can one degrade food into small pieces.

In my study, I collected 60 people with malocclusion (30 males, 30 females), and then matched with them 60 controls (according to age and gender) without malocclusion (30 males, 30 females).

Then I checked whether the indicators of "mastication performance" differed significantly between the individuals without malocclusion (control) and those with malocclusion (case? cohort?).

Yes, as you say, I collected the people from the same place. Then checked whether they had malocclusion or not, and checked what the "mastication efficacy" rate was in each group. Then as you said it is a retrospective cohort.

A VERY BIG "THANKS" FOR CONFIRMATION.

My problem was the technical definition of "malocclusion". Although it seems to be the exposure here, it is a disorder in nature. So my statistician keeps on saying it is a case-control study!

My data are summarized in a table:

There are two columns: with malocclusion, without malocclusion
each column have a "mastication performance indicator" value. Its unit and nature is too technical (dentistry) to be described here, but those with higher values of this indicator are likely to be poorer on mastication.

I compared these values using a t test and found that there is a significant difference between the mastication efficacy of those groups (control vs. case? / cohort?).

Are you still sure it is a (retro) cohort?

##### Ninja say what!?!
I can see why the statistician thinks it is a case control study.

Let me make sure that I got this straight though. In your study, you collected 60 people with the exposure of interest. You then matched 60 people without the exposure. After doing so, you then checked whether the disease rates differed among the groups. Am I correct?

If you agree with what I just said above, then yes you do have an age and gender matched retrospective cohort design. The statistician seems to think you have a case control because you've matched on a disease. However, you are correct here in considering the matching factor to be the exposure. The groups that you've matched on is the exposure of interest and not the outcome.

Hope that helps.

PS. If you're interested in having me check your answers, please post the 2x2 table that you constructed. I'll take a look at the numbers and tell you what I think.

#### victorxstc

##### Pirate
Your help means a LOT to me. You just saved me

Yes, exactly as you said. I collected those with the exposure (which is a disease in nature) and matched 60 controls and then compared the rates of the mastication efficiency of the two groups.

I really appreciate your kind help.

About the table, it is as follows (although the statistician has determined the test as t, due to the sample distribution etc.):

--------------------------------Malocclusion ----------------------- Non-malocclusion (control)

Mastication performance --- 2.91 +- 0.38 ---------------------- 4.74 +- 0.52 -----------> P < 0.05

Last edited:

#### victorxstc

##### Pirate
When I go much deeper in detail, I found this study partly case control, and partly cohort.

The cohort part is what you did tell me.

Another part is that a table exists which investigates the association between age, gender, and stature with malocclusion itself. In this new part, malocclusion is not the exposure, but is an outcome.

But the table is divided again based on the malocclusion (which is the outcome here), not the independent variables age, gender, and stature. (although stature can be both dependent and independent, because it can be affected by malnutrition, and can affect the mastication too).

I don't know how can I send the original tables to you, but if you don't mind I would be so thankful.

But my problem right now is that is it possible to design and run a study which is both case control and cohort?!

the other table is like this (I skipped the values):

------------------------malocclusion (here it is really the outcome)-------control

Female--------age:-----mean and SD (in year)-------------------------- mean SD ---> P value
-------------weight:-----mean and SD (in Kg)---------------------------- mean SD ---> P value
-------------Stature:----mean n SD (in mm)------------------------------ mean SD ---> P value

Male-----------age:-----mean and SD (in year)-------------------------- mean SD ---> P value
-------------weight:-----mean and SD (in Kg)---------------------------- mean SD ---> P value
-------------Stature:----mean n SD (in mm)------------------------------ mean SD ---> P value

I just don't know what should I write in my paper, that "this was a combination of cohort and case control designs?" Do you know is it correct to say that your study is both case control and cohort?

Thanks a lot

Last edited:

#### victorxstc

##### Pirate
I think it might be a retrospective cohort study with a "nested case control" study.

##### Ninja say what!?!
I think it might be a retrospective cohort study with a "nested case control" study.
I don't know that it's really this. Usually, it's really only one or the other. Do you have an epidemiologist that you can come to?

#### victorxstc

##### Pirate
Thanks a lot Link. My statistician is an epidemiologist (PhD)! I think this study is both. I think this is the fault of the statistician to analyze the first part as a case control study, and analyze the second part as a cohort. Apparently he came to split all the data according to the malocclusion factor (which was the outcome in the first part [associations with age, height, etc.] but was the exposure in the second part [association with mastication performance]).

He could easily split the sample in the first phase according to the demographic data (age, etc.) not the malocclusion, to check the causation as well. Unfortunately he has truncated the raw data.

I appreciate all your kindness.

Last edited: