Case-control vs. Cohort studies

The goal of most outbreak investigations is to identify the source of the outbreak. In some outbreaks, information available at the outset of the investigation clearly supports a particular source, and all that is needed is to carefully examine the established facts. In the majority of outbreaks, however, the circumstances are not that straightforward, and an analytic study needs to be undertaken.

The gold standard for analytic studies is an experimental study design. In an experimental study, investigators try to change the disease outcome among parti- cipants by manipulating the exposure while holding all other variables constant (e.g., a clinical drug trial). In the investigation of outbreaks, however, investigators cannot manipulate exposures. (Exposures have generally already occurred. Furthermore, because most exposures of interest in an outbreak might have a negative impact on the subject, assignment of exposure would be unethical.) Therefore, analytic studies of outbreaks are almost always observational.

In an observational study, the investigator observes the course that nature has taken. The investigator selects a group of subjects who already have the expo- sure of interest or the disease of interest and compares this group to another group (the comparison group) whose members have not had the exposure or experienced the disease. The investigator then draws conclusions about the association between the exposure and the disease by comparing the two groups.

In outbreak investigations, two types of observational studies are commonly used, case-control and cohort studies. We will talk about each of these study types.


After completing this lesson, you will be able to...

  • describe the two types of analytic studies routinely used to investi gate outbreaks
  • list the strengths and weaknesses of each study type
  • given the details of a particular outbreak, decide which type of analytic study to undertake

Case-control study
In a case-control study, subjects are enrolled on the basis of whether they have (or had) the disease of interest or not. Previous exposure to the factor of interest is then compared between those with (i.e., cases) and without the disease (i.e., controls) to see if an association between disease and exposure exists.

Case-control studies are relatively economical and quick. They are ideal for studying rare diseases or diseases that can be defined specifically. (Inves tigators can pointedly search for persons with the disease and enroll them into the study.) Because entrance into a case-control study begins with disease status, investigators can examine the association between multiple exposures and the disease of interest. Case-control studies typically provide greater statistical power than cohort studies (of the same size) to detect differences (if they exist). Finally, case-control studies are particularly appropriate when the population at risk is ill-defined or difficult to distinguish from the general population.

Design of a Case-control Study

Because entrance into a case-control study is based on disease status, investigators cannot examine multiple diseases that might be associated with an exposure. If the population at risk is ill-defined, selection of appropriate controls can be challenging. Finally, the data collected in a case-control study normally cannot be used to directly calculate absolute risk for disease after an exposure.

Cohort study
In a cohort study, subjects are enrolled on the
basis of exposure
(or lack of exposure) to a factor
or on membership in a well-defined population. Occurrence of disease is then compared between persons with and without the exposure to determine if an association between exposure and disease exists.

Cohort studies can be prospective or retrospective. In a prospective cohort study, participants are enrolled before exposure occurs. In a retrospective study, the study begins after exposure occurs. For outbreak investigations, only retrospective studies are used.

Cohort studies are well-suited for examining disease after rare and highly specific exposures. (All persons with the exposure can be enrolled into the study.) Because entrance into a cohort study begins with exposure status, investigators can monitor the occurrence of multiple diseases potentially caused by an exposure. Cohort studies allow the direct measurement of the absolute risk of developing a disease after an exposure. Finally, cohort studies are

Design of a Cohort Study

particularly appropriate when a small, well- defined group has been exposed to a disease agent (e.g., guests at a wedding).

Because entrance into a cohort study is based on exposure status, cohort studies can examine only a single exposure or group of related exposures. In addition, cohort studies cannot be used if the population at risk is not known (i.e., as with nationwide epidemics).

When to use each study type
To say the least, it is often not easy to decide which study type to use in an outbreak investigation. This table summarizes the strengths and weakness of each study type and when using each one is best.

Comparison of Study Types
Study type Strengths Weaknesses When best to use
Case-control study
  • economical and quick
  • good for study of rare or specifically defined diseases
  • can examine multiple exposures
  • most statistical power
  • can be used if population at risk is unknown
  • cannot examine multiple diseases
  • difficulty in selection of comparison group
  • cannot be used to calculate absolute risk
  • when disease is rare or can be defined with great specificity
  • when exposure is common
  • when the population at risk is unknown
Cohort study
  • good for study of rare or specifically defined exposures
  • can examine multiple diseases
  • provides most direct measurement of absolute risk
  • can study only single exposure or group of related exposures
  • population at risk must be known
  • when the population at risk is a well-defined group
  • when exposure is rare or can be defined with great specificity
  • when disease is common

Questions to Guide Selection of Study Type

  • What is already known about the exposure-disease
  • How common is the disease?
  • How common is the exposure?
  • Do you have access to a study population?
  • What budget/resources do you have?
  • How pressing is the need to control the disease?

If you are still unsure about the best study type, the following questions might also help:

What do you already know about the exposure- disease association? If little is known (and multiple exposures are of interest), you probably want to do a case-control study. If you already know that a disease affects persons with a specific exposure, you probably want to do a cohort study.

How common is the disease or exposure of interest? If the disease is rare, you would probably do a case- control study and enroll all known case-patients. If the exposure is rare, you would probably do a cohort study and enroll all known exposed persons.

Do you have access to a particular study population? If you have access to cases through a disease regis- try, a case-control study might make more sense. If a registry of persons with the exposure exists, a cohort study might make more sense. If the problem affects a small well-defined population, a cohort study might make most sense.

What resources can you use to do the study? Because of the greater study power, a case-control study might allow you to limit the number of study participants and decrease necessary resources.

How pressing is the need to control the disease? The urgency of the situation might govern your decision. Conducting a case-control study within a large, well-defined cohort might be more appropriate if it will give an answer more quickly.

As you can see, deciding on the study type is not always straightforward. Give it a try with the following optional exercises.

Exercise 1

In 1989, a new disease, eosinophilia myalgia syndrome (EMS), was recognized after physicians in New Mexico and Minnesota examined three patients with severe muscle pain and a marked increase in eosinophils. All three patients had been taking a food supplement called L-tryptophan. An announcement about the cases led to reports of over 250 additional cases of EMS from 37 states.

Would you undertake a cohort or case-control study to investigate this outbreak?
Click here to see answer.

The population at risk for this new illness is not known. With <300 cases, it is also relatively rare. Although exposure to L-tryptophan is of interest, it is only one of many exposures you probably want to examine. And because the outbreak appears to be ongoing, you want a quick answer. Therefore, a case-control study is probably your best bet.

In the actual investigation …
Health departments in New Mexico and Minnesota each conducted a case-control study. In the New Mexico study,

12 EMS cases were compared with 24 matched controls. All case-patients and two (8%) controls had used L-tryptophan (odds ratio [OR] not calculable; p = 0.0000069).

In Minnesota, 12 cases were compared with 12 matched controls. All case-patients and no controls had used L-tryptophan during the month before illness onset (OR not calculable;
p = 0.0008).

EMS was eventually linked to one specific brand of L-tryptophan manufactured in Japan. It is believed that a contaminant unique to this brand, not the L-tryptophan itself, played a role in the development of EMS among users.

Source: CDC. Epidemiologic notes and reports eosinophilia-myalgia syndrome and L-tryptophan-containing products -- New Mexico, Minnesota, Oregon, and New York, 1989. MMWR 1989;38(46);785-8.

Exercise 2

On May 20, 1996, the Ontario Health Department notified the Texas Department of Health (TDH) about Cyclospora infections in three Canadian businessmen who had attended a meeting at a private club in Houston, Texas. Twenty-eight persons had attended the business meeting.

Would you undertake a cohort or case-control study to investigate this outbreak?
Click here to see answer.

The cyclosporiasis outbreak in Texas was associated with a business meeting that involved a small, well-defined group of attendees. Therefore, use of a cohort study to identify the source of the outbreak seems reasonable. (Because the study was initiated after the exposure leading to the outbreak, it would be a retrospective cohort study.)

In the actual investigation …
TDH undertook a retrospective cohort study to investigate the source of the outbreak.

Investigators questioned all persons who attended the meeting to determine if they had cyclosporiasis and to collect information on recent travel and food and water exposures at the meeting.

Twenty-four meeting attendees provided information. Attendees who ate a mixed berry dessert served at the meeting were more likely to have become ill than those who had not eaten the dessert. No other exposures were associated with illness.

At the same time as the Texas outbreak, cyclosporiasis cases increased nationwide. Studies in other states linked the outbreak to Guatemalan raspberries. Later, raspberries were identified as a component of the mixed berry dessert served at the Houston business meeting.

Source: Texas Department of Health. Outbreaks of Cyclospora infection. Disease Prevention News 1997;57(11):1-4.

Exercise 3

On September 28, 2002, a cruise ship embarked with 1,984 passengers and 941 crew members for a 7-day round-trip cruise from Florida to the Caribbean. By October 1, at least 70 passengers and two crew members reported a gastrointestinal illness suggestive of norovirus infection. On October 3, CDC epidem- iologists boarded the ship to conduct an investigation. Passengers were due to disembark on October 4.

Would you undertake a cohort or case-control study to investigate this outbreak?
Click here to see answer.

You could conduct either a cohort or a case-control study to investigate this outbreak. The population at risk was well-defined (i.e., cruise ship passengers and crew) and easily accessible, supporting the choice of a cohort study. With the large number of subjects, however, a cohort study might take substantial time. (A random sample of the cohort could be undertaken. However, a sample would decrease the number of cases included and, therefore, the power of the study.) A case-control study would allow investigators to limit the number of

subjects in the study, include sufficient numbers of cases to ensure adequate study power, and complete the study more quickly. As a result, a case-control study might be advantageous.

In the actual investigation …
Investigators conducted a retrospective cohort study. The study was quickly accomplished by distributing self-administered questionnaires to every ship cabin. Questionnaires completed by 1,879 passengers identified 356 cases of acute gastrointestinal illness. Passengers who ate lunch served at embarkation were more likely to become ill than passengers who did not (relative risk = 2.4; p-value = 0.02). Secondary cases occurred as a result of person-to-person spread.

Source: CDC. Outbreaks of gastroenteritis associated with noroviruses on cruise ships --- United States, 2002.
MMWR 2002;51(49);1112-5. .
Page  of