Yes, you can design case control study because you have the outcome and search the exposure our risk factor

Dear Tauqeer Hussain Mallhi,

Some questions:(1) From which date on and until what date have you collected data for each patient (eg, from dengue diagnosis to death or hospital discharge?)? Also historical data about conditions before the date of inclusion into study?

(2) How many of those 700 patients were followed-up since date of inclusion into the study and for how long? How long was the median follow-up, how long for each of those 8 who died?  (Dengue-associated death may occur within days, right?)

(3) What's your exposure of interest? Is it a time-varying exposure that may change during your follow-up or was it fixed since inclusion of patients into study (until death)?

Since you have only 8 cases, the possibilities to assess risk factors are limited. In fact, if none of your 8 cases was actually exposed to the exposure of interest then you might not be able to calculate a meaningful risk estimate at all. 

Kind regards, René

@M Hammad thank you to build my confidence

Respected Rene thank you so much for your concern.... you asked three main questions and answers of them are below

1. I collected data of dengue patients admitted to our hospital from 2008 to 2014. For each patient I collected data on demographics and laboratory characteristics from date of admission until discharge.... It means death occurred during hospital stay due to dengue infection.

2. We did not follow up these patients, as we waited for them to be admitted to our hospital, its just like observational study. then they admitted to hospital and we recorded data for each day of hospital stay until discharge or death, whichever occurred early.

3. Actually my interest is to study clinical and laboratory characteristics of dengue viral infection in my hospital therefore I also planned to study fatal cases. As fatal cases are very few in my study so it will be inappropriate that I compared 8 death cases with 692 non-death cases. Therefore I am thinking to design a case control study rather than case series study.

I know that by using this technique, statistical power of risk factors will be decreased. I found a paper of singapore, they did by same way as i mentioned. They took 3000 patients appx. and then studied 7 fatal cases against 35 non-fatal cases. 

Thank you Sofia, I ll go through it

The number of Controls (1x40) will increase the statistical Power of your study. However, as important as the level of significant of the study is to control so much as possible your confounders (age, sex, occupation, level of education, residence, admission time, severity, other Health problems, etc).

@Richard sir, thank you, you built my confidence on my idea

Yes you can do a case control study. After 6 control subjects the power that you increase your study is very small. Read Schlesselman case control studies.

Yes you can apply a case control study design. The only thing you should be cautious is ascertainment of controls. Make sure that the cases are with the outcome (at least fulfill the cased definition) of interest and controls are without the outcome during the study period. The power of the study has to be also reasonable 80% or more when calculating your sample.

Yes, it is not easy because, of problem of case and control group (selection)

and comparability and not easy to explaine and interpretate the result

Note; you can return to( Cole 1979)

Of course you can do it but have two options: a) Comparing dead cases to non dead ones (even all of them) without a working hypothesis -in what is called in  epidemiological grounds a "fishing expedition",  or b) By formulating a working hypothesis that will allow you to specify the sample size (both your already collected cases and the "n" controls needed) and trading off the number of matched  controls needed. In this case, your working hypothesis needs stand that a certain factor is associated to death. For such a proposal the hypothesis needs to be sound, something that might be difficult to support. I would give you the following advice: Carry out the first operation -the fishing expedition- for futther careful examination of the results, and then try to identify a reasonable exposition factor as associated to. Afterwards, make up a true case-control design, with an explicit hypothesis. Depending of the problem nature, eight cases maight be good enough, provided you find out a reasonably "good" risk factor.

I am attaching the following reference to enhance your understanding of how it may be done, although the situation is not exactly the same. I already warn you about the number of controls: the study power will not increase ad infinitum by freely increasing controls. Actually, besides of multiple controls, what you need is matching controls, as somebody says above.

Te reference I mention is:

Herbst AL, Ulfelder H, Poskanzer DC, “Adenocarcinoma of the vagina: association of maternal stilbestrol therapy with tumor appearance in young women”. NEJM 1971;284:878-81.

Good luck!.

Dr Patricio thank you so much for you guidance, i ll go through the reference you gave

The design of your study could be that of a case-control study nested within your cohort of dengue patients. But Rene Schade's remarks should be taken into account. With only 8 cases, even if the number of controls is multiplied by 5, your study is likely to not be powerful unless you want to detect huge odds ratios. This is not usually the case.

I wuold agree with Abdeldjellil Bezzaoucha  about his statement about ORs. Yet, if it happens, decissions have to be made about whether a large OR - if fobtained- makes or not biological sense.  In Tauqeer's situation, the question becomes: Should Tauqeer Hussain start his resarch right now or wait to more death cases?. Following my oprevious opinions I would suggest him to start it soon, the new "cases" being added in a nested case-control design. This may be done provided careful matching (randomly) is made whenver a case emerges. Nonetheless, what is important is to learn about the time related expectation of new "cases".

There must be a misunderstanding. The nested case-control y mention would consist on adding new cases -an their concurrent controls,. Of course this study, as it is now,  cannot be taken as a nested case-control.

I am agree with Tahir and Patricio, nested means to include new cases with time. I have already collected data of 700 patients and among them 8 are fatal cases

The inclusion of new cases when performing a case-control study makes this case-control study a prospective study (prospective case-control study). Do not confuse the study structure with the chronology of data collection.

A nested case-control study only means that controls are selected from the same cohort that generated the cases (new cases or no).

In reality, your problem, at this stage,  is essentially the preliminary calculation of sample sizes  needed to achieve your study (eg use the free calculator OpenEpi).

The sample size calculation is almost a nonsense issue if a working hypothesis is not properly stated. In this case, the hypothesis would include -as an example- the exposition rate in controls (requiring  a hypothetical factor) and the hypothesis expected OR, besides the other components of the case-control proposal design: erros I and II level accepted for the calculations.

FIRST, YOU MUST KNOW WHAT KIND OF KNOWLEDGE NEEDS TO DEVELOP.

IF YOU WANT TO KNOW ABOUT MORTALITY IN DENGUE, IS ADVISABLE A MORTALITY STUDY.

Dear Tauqeer Mallhi,

What kind of information did you want to generate or hypothesis to test? That has to be clear. If it is hypothesis generation, , why don't you go for  descriptive epidemiological study designs like case series as you have a small number of cases to deal with. 

Dear friends

The question of the number of subjects required to include in a study is a matter of substance . It must specify the interesting difference that one wishes to demonstrate . This involves detailed knowledge of the subject and a very good understanding of the issues.

For a case control study, You must specify the desired confidence level (1 - error I), power (1 - error II), a hypothetical percentage of exposure (for an important factor) among the controls, and either an odds ratio or a hypothetical percentage of exposure among the cases :

                                                               www.openepi.com

www.openepi.com

Dear friend,  The most vital thing to remember is that case control studies are to test the hypothesis. in your case what are you interested to prove or disprove in dengue patients.. if you have some hypothesis is with you , pl . let me know , i shall guide you , how to proceed. Are you interested to know the role of some factor causing death in dengue patients?

@Ijaz Ahmad, Sir thank you so much for your concern.... Yes you are right. My concern is to evaluate characteristics between fatal and non-fatal cases and to determine risk factors of mortality in dengue patients.

In a nested case-control,  the inclusion of controls (new) at the time new cases are added is essential. When you do it so, you are matching the new cotrol(s) to each new case by chronology. This is the only way to cope with risk of biases the case-control has as a design. The other advise I would ask you to take is to not use softwares for sample size calculation unless you have a hypothesis, and hence, all de values required to perfom de calculations (i.e. the size of the difference of the studied variable in cases and controls as expected from the hypothesis analysis. This difference has to be expressed either as an OR or a RR, the first being the one that corresponds to a case-control study.

Dear malhi , very easy . You spell out the factoror /factors you want to compare.between fatal and non fatal cases. This step will clear up your mind and one can guide you to proceed further.

The first issue for Tauqueer is to state which is (are) the factor associated to death. As I understand, he has not identified one or more but rather, what he wants is to look for those with highest degrees os association. If there is no one shown by the literature -as an example- he must start just comparing his 8 cases to the remaining 692. (That is what I mean by "fishing expedition"). If one or more factors appear as associated to death, he should select the one that will make up his hypothesis with. I would suggest to pick -among those with statistical significance- that which makes biological sense. If the software used for this first analysis yield some statistical significant association measure, I would pick -among them- the lower value (i.e. the lower OR). The explanation for this would take much more room and time.

I guess your objective will be to establish the risk factors associated with mortality in a group of patients suffering from dengue fever. i don't like seeing challenges..but my friend you have many challenges here..To start with the number of deaths from your group is only 8 out of 700 patients...so even if you were to have all deaths as cases and all survivors as controls you only have 8 subjects..against 692 controls..its practically impossible to randomly selected cases..and therefore you will have a problem explaining that any observation in the cases occurred by chance. This is not to say that you cannot do a paper from your data..you can use a descriptive methodology to write your paper..because it will provide critical knowledge to the world...secondly: your data collection fits a Prospective cohort design..but lacks a control group. so to start with in 2009..you have been following cases only whose exposure status is dengue fever..therefore its practically impossible to do a case control: Its possible to do a case control study within  a Prospective cohort study..How? in the design stage you need to identify 2 groups of people based on exposure status..exposed group and non exposed group..then you need to follow them up until they develop disease...then you can randomly choose cases from these who have developed the disease and Controls from these without the disease..then by use of regression analysis..you can quantify the risks associated with development of the disease in question..such will be a Nested Case Control Study..Nested because its a study within a study..Thanks..Nzioki-Mativo

This exchange started with Tauqeer's question: whether a case-control design is possible given his data. One of the reasons the colleagues have sent their contributions -included myself- is that death is -in his series- an uncommon event, This is in itself a good reason to discuss a case-control design. A cohort design will take a long portion of  Tauqeer's life!.  

@Patrico Sir, Thank you so much. lolll for long portion of life. No doubt, I am not predictable about duration of my life so better I start with case control study.

Were your controls tested for dengue? 

Yes, case control study can be conducted . But controls must be tested to be free of dengue . Secondly what risk factors are under study . OR author is interested in " fishing expedition and want to determine the associated risk factors ? 

nice information

following answers