A cohort study takes a group of individuals and usually follows them forward in time, the aim being to study whether exposure to a particular aetiological factor will affect the incidence of a disease outcome in the future (Fig. 15.1). If so, the factor is generally known as a risk factor for the disease outcome. For example, a number of cohort studies have investigated the relationship between dietary factors and cancer. Although most cohort studies are prospective, historical cohorts are occasionally used: these are identified retrospectively and relevant information relating to outcomes and exposures of interest up to the present day ascertained using medical records and memory. However, while these studies are often quicker and cheaper to perform than prospective cohort studies, the quality of historical studies may be poor as the information collected may be unreliable.

Figure 15.1 Diagrammatic representation of a cohort study (frequencies in parentheses, see Table 15.1).

Cohort studies can either be fixed or dynamic. If individuals leave a fixed cohort, they are not replaced. In dynamic cohorts, individuals may drop out of the cohort, and new individuals may join as they become eligible.

Selection of Cohorts

The cohort should be representative of the population to which the results will be generalized. It is often advantageous if the individuals can be recruited from a similar source, such as a particular occupational group (e.g. civil servants, medical practitioners), as information on mortality and morbidity can be easily obtained from records held at the place of work, and individuals can be re-contacted when necessary. However, such a cohort may not be truly representative of the general population, and may be healthier. Cohorts can also be recruited from GP lists, ensuring that a group of individuals with different health states is included in the study. However, these patients tend to be of similar social backgrounds because they live in the same area.

When trying to assess the aetiological effect of a risk factor, individuals recruited to cohorts should be disease-free at the start of the study. This is to ensure that any exposure to the risk factor occurs before the outcome, thus enabling a causal role for the factor to be postulated. Because individuals are disease-free at the start of the study, we often see a healthy entrant effect. Mortality rates in the first period of the study are then often lower than would be expected in the general population. This will be apparent when mortality rates start to increase suddenly a few years into the study.

Follow-Up of Individuals

When following individuals over time, there is always the problem that they may be lost to follow-up. Individuals may move without leaving a forwarding address, or they may decide that they wish to leave the study. The benefits of a cohort study are reduced if a large number of individuals is lost to follow-up. We should thus find ways to minimize these drop-outs, e.g. by maintaining regular contact with the individuals.

Information on Outcomes and Exposures

It is important to obtain full and accurate information on disease outcomes, e.g. mortality and illness from different causes. This may entail searching through disease registries, mortality statistics, and GP and hospital records.

Exposure to the risks of interest may change over the study period. For example, when assessing the relationship between alcohol consumption and heart disease, an individual’s typical alcohol consumption is likely to change over time. Therefore it is important to re-interview individuals in the study on repeated occasions to study changes in exposure over time.

Analysis of Cohort Studies

Table 15.1 shows observed frequencies. Because patients are followed longitudinally over time, it is possible to estimate the risk of developing the disease in the population, by calculating the risk in the sample studied.

Table 15.1 Observed Frequencies (see Fig. 15.1)

The risk of disease in the individuals exposed and unexposed to the factor of interest in the population can be estimated in the same way.

Estimated risk of disease in the exposed group,

Estimated risk of disease in the unexposed group:

Then,

The relative risk (RR) indicates the increased (or decreased) risk of disease associated with exposure to the factor of interest. A relative risk of one indicates that the risk is the same in the exposed and unexposed groups. A relative risk greater than one indicates that there is an increased risk in the exposed group compared with the unexposed group; a relative risk less than one indicates a reduction in the risk of disease in the exposed group. For example, a relative risk of 2 would indicate that individuals in the exposed group had twice the risk of disease of those in the unexposed group.

A relative risk should always be interpreted alongside the underlying risk of the disease. Even a large relative risk may have limited clinical implications when the underlying risk of disease is very small.

A confidence interval for the relative risk should be calculated, and we can use it or determine a test statistic to test the null hypothesis that the true RR = 1. These calculations are easily performed on a computer and therefore we omit details.

Advantages of Cohort Studies

• The time sequence of events can be assessed.
  
• They can provide information on a wide range of disease outcomes.
  
• The incidence/risk of disease can be measured directly.
  
• Very detailed information on exposure to a wide range of factors can be collected.
  
• They can be used to study exposure to factors that are rare.
  
• Exposure can be measured at a number of time points in each study, so that changes in exposure over time can be studied.
  
• There is reduced recall and selection bias compared with case–control studies (Chapter 16).

Disadvantages of Cohort Studies

• In general, a cohort study follows individuals for long periods of time, and it is therefore costly to perform.
  
• Where the outcome of interest is rare, a very large sample size is required.
  
• As follow-up increases, there is often increased loss of patients as they migrate or leave the study, leading to biased results.
  
• As a consequence of the long time-scale, it is often difficult to maintain consistency of measurements and outcomes over time. Furthermore, individuals may modify their behaviour after an initial interview.
  
• It is possible that disease outcomes and their probabilities, or the aetiology of disease itself, may change over time.

Study Management

Although cohort studies are usually less regulated than clinical trials (Chapter 14), it is still helpful to prepare a study protocol at the outset of any cohort study. It is important to pay particular attention to the following aspects of study management when preparing this document.

• The outcome of interest: specify the outcome (e.g. obesity) and provide an unambiguous definition of it (e.g. body mass index > 30 kg/m²). How will it be ascertained (e.g. through direct contact with patients, through access to hospital records or through linkage with national registries)?
  
• The exposures of interest: specify which exposure variables will be considered and give unambiguous definitions of them. How will the exposures be ascertained?
  
• Monitoring of participants: how will participants be monitored (e.g. by direct face-to-face visits, through postal questionnaires, through access to hospital records)? How frequently will participants be followed up? What information will be collected at each time point? Will any biological samples (e.g. blood, urine, biopsy samples) be collected?
  
• The size of cohort and length of follow-up: how frequently is the outcome likely to occur in those with and without the exposures of interest? How ‘big’ should the study be to ensure that the study is sufficiently large to demonstrate associations of interest? Note that in a cohort setting, the power of a study (Chapters 18 and 36) is largely determined by the number of events that occur; this can be increased either by increasing the size of the cohort or by lengthening the period of follow-up.
  
• The definition and ascertainment of any potential confounders (Chapter 34) and/or effect modifiers: specify which other important variables should be investigated and provide an unambiguous definition for each.
  
• The plans for statistical analysis: when is it anticipated that the statistical analysis of the cohort will be undertaken (e.g. after five years)?
  
• The steps taken to reduce bias (Chapter 34): what steps will be taken to minimize drop-out from the cohort? What steps will be taken to ensure that the definition and ascertainment of outcomes, exposures and other key variables do not change over time?
  
• The plans for quality control: describe any statistical analyses that will be conducted at interim time points (Chapter 18) to ensure that:
  
  
  
  • loss-to-follow-up is not substantial;
    
  • the way in which exposures, outcomes and other key data are measured or ascertained has not changed over time; and
    
  • outcomes are occurring at the rate expected at the outset such that the study is ‘on target’ for the planned analyses.
  
  
• The need for ethics committee approval and/or patient consent: will these be required? If patient consent is required, how will this be collected?

Clinical Cohorts

Sometimes we select a cohort of patients with the same clinical condition attending one or more hospitals and follow them (either as inpatients or outpatients) to see how many patients experience a resolution (in the case of a positive outcome of the condition) or some indication of disease progression such as death or relapse. The information we collect on each patient is usually that which is collected as part of routine clinical care. The aims of clinical cohorts (sometimes called disease registers or observational databases) may include describing the outcomes of individuals with the condition and assessing the effects of different approaches to treatment (e.g. different drugs or different treatment modalities). In contrast to randomized controlled trials (Chapter 14), which often include a highly selective sample of individuals who are willing to participate in the trial, clinical cohorts often include all patients with the condition at the hospitals in the study. Thus, outcomes from these cohorts are thought to more accurately reflect the outcomes that would be seen in clinical practice. However, as allocation to treatment in these studies is not randomized (Chapter 14), clinical cohorts are particularly prone to confounding bias (Chapter 34).

Stay updated, free articles. Join our Telegram channel

Join