<p>After going through this article reader must be able to understand: Basic design issues of cohort studies, strengths and weakness of cohort study design, analysis of cohort studies and critical appraised of cohort studies encountered in medical literature</p></sec><sec id="sec001"><title>Introduction

Randomized control trials are considered the best method of generating evidence on efficacy of an intervention in a highly selected population. Cohort studies can be thought of as natural experiments in which outcomes are measured in real world rather than in experimental settings (1). ‘Cohort’ was the Roman term for a group of soldiers marching together. A cohort is a group of people who have something in common and who remain part of a group over an extended period of time. Subjects in cohort studies are usually selected because they have some characteristic or some kind of occurrence (exposure) which places them at higher or lower risk of having an outcome (disease).

Cohort studies ask the question “what will happen?” Therefore the direction in cohort studies is forward in relation to time. At the onset of the study researchers select a group of subjects and determine whether they have risk factors (exposure). The subjects, both exposed & unexposed are followed up for a specified period of time to observe the frequency or type of outcome (Figure 1). While selecting cohort for a particular outcome, the subjects selected should be free from outcome of interest, but should be susceptible to the outcome of interest.

Cohort studies serve two primary purposes: descriptive (to describe the incidence of certain outcomes over time) and analytic (to analyze association between predictor and outcome). For example an investigator selects a cohort of babies with birth weight less than 1000 gm (exposure) and studies the incidence of cognitive disabilities (outcome). Investigator can also compare the outcome in this group of babies (exposed) with outcome in babies with normal birth weight to analyze the risk due to exposure status (birth weight < 1000 gm).

Strengths and weaknesses

Strengths

Powerful strategy for defining the incidence and investigating the potential causes of a condition.

Can establish timing and directionality of events.

Forward study design allows to measure important variables completely and accurately.factors of one outcome of interest.

Can examine multiple effects of a single exposure

Weaknesses

Expensive

Inefficient for rare outcomes Potential for selection bias

Vulnerable to loss of follow up (attrition bias)

Long duration when done Prospectively

Can study few exposures

Planning a cohort studySelection of subjects

Contrary to randomized controlled trials, in cohort studies allocation of intervention to subjects is not by chance. Therefore cohort studies are vulnerable to selection bias and care must be taken to minimize, assess and deal with selection bias. Ideally comparison group in the cohort study should be identical to the exposed (intervention) group apart from the fact that they have not been exposed to the risk factor. The subjects should be appropriate to the research question and available for follow up. They should sufficiently resemble the population to which the results will be generalized. Subjects should be free from the outcome of interest. This may require application of a sensitive screening test to rule out the sub-clinical outcome (disease).

Adequate sample size

The number of subjects should provide adequate precision and power.

Accurate measurement

The quality of the study will depend on the precision and accuracy of predictor and outcome variables. Following steps help in minimizing measurement bias

Use pre-decided diagnostic criteria

Identical procedures for disease identification for both groups

Wherever possible, observers blinded with regard to exposure should assess outcome

Methods of measurement of exposure identical in both groups

Adequate follow up

Steps should be taken to minimize the loss of follow up. Best is to have follow up rates up to 90%. Following steps should be taken to minimize the loss of follow up:

During enrolment:

Exclude those likely to be lost (planning to move or unwilling to return)

Obtain information to allow future tracking (address, telephone number of subjects and close relatives)

During follow up:

Periodic contact with subjects

Analysis of cohort study

Results from cohort studies provide two types of information: measures of disease frequency and measures of association.

Measures of disease frequency

Measures of disease association

Cohort study and role of confounders

Confounding is a situation in which the estimated intervention effect is biased because of some difference between the comparison groups apart from the planned interventions such as baseline characteristics, prognostic factors or concomitant interventions (2). For a factor to be a confounder, it must differ between comparison groups and predict the outcome of interest. In randomized controlled trails all potential confounders are expected to be evenly distributed between groups being compared. Cohort studies however have no similar protection against confounding and are especially vulnerable to unknown confounders. Therefore all efforts must be made at the design stage itself to identify all the potential confounders. Two analytic strategies-regression and stratification can be used to assess and reduce confounding. Stratification is the process in which the sample is divided into subgroups or strata so as to balance potential confounder(s). Regression uses the data to estimate how confounders are related to the outcome and produces an adjusted estimate of the intervention effect. It is the most common method used for reducing confounding in cohort studies.

Cohort studies in Indian literature

Paul V K et al. followed up a cohort of high risk neonates requiring intensive care to determine the neuro-developmental outcome. One hundred and nineteen nursery graduates were enrolled for follow up if they fulfilled any of the following risk factors: birth weight less than 1500 g, Apgar score less than 4 at 5 minutes, seizure(s), and required assisted ventilation for more than 24 hours. They included 55 infants with birth weight < 1500 g, 45 with low Apgar scores, 12 with seizure(s) and 28 who received assisted ventilation. An overwhelming majority of subjects (85%) had normal neuro-developmental outcome. The adverse outcome in the remaining 15 included mental retardation in all, subnormal motor development in 14, microcephaly in 1, hearing loss in 2 and visual impairment in 4. Among the neonatal risk factors, seizures, sepsis and hypoxic ischemic encephalopathy had a significant association with adverse outcome (3).

Chaudhari S et al. studied a cohort of babies with birth weight less than 2000 grams and reported their cognitive at the age of 12 years. One hundred and eighty children weighing less than 2000 grams at birth and ninety control children were assessed. The intelligence and academic performance of the children weighing less than 2000 grams was found to be significantly lower, though within normal limits. These babies also had poor visuo-motor perception, motor incompetence, reading and mathematics learning disability. The preterm SGA and VLBW children had the poorest cognitive abilities (4).

Bhargava S K et al. studies the physical growth and sexual maturation of children born with low birth weight (< 2000 g). and found that fetal growth retardation has a lasting adverse effect on later physical growth (5).

Checklist for critical appraisal of a cohort study

Is cohort study design the right method of choice for asking the research question investigator is seeking to answer?

What are the comparison groups? What comparisons are being made and are these comparisons valid?

What are the potential selection biases? What steps have been taken to minimize the selection bias?

Has there been a systemic effort to identify and measure potential confounders?

What steps have been made to deal with confounders?

Are the analytic strategies clearly described?

Are the results plausible?

Design and conduct of cohort studies