Designing Cohort Studies

Cohort Studies
An incidence study is a subtype of longitudinal study in which the outcome measure is dichotomous (e.g. death or disease incidence). Perhaps the simplest type of incidence study involves “descriptive” analyses using routine mortality or incidence records for a defined geographic population. For example, most countries have comprehensive death registration schemes, as well as regular national censuses, a population register, or other methods of estimating population numbers. These can then be used, as the numerator and denominator respectively, to calculate overall national death rates, as well as the death rates by age-group and gender. In some countries, information may also be available to calculate death rates by other demographic variables such as ethnicity, socio-economic status, employment status, occupation or geographical area. However, the validity of such analyses may be questionable, because in most countries death certificates (or other routine records such as cancer registration records) are not linked directly to the corresponding population records. Thus, problems may occur if factors such as ethnicity are coded differently on the death records and on the population records. Nevertheless, such “descriptive” analyses, have played a major role in identifying public health problems and suggesting priorities for public health research.
However, the limitations of analyses based on routine records usually mean that a specific “cohort” must be constructed for many epidemiologic studies.
Defining the source population and risk period
Community-based cohort studies
For studies investigating environmental factors, or general lifestyle (diet, exercise, etc) a cohort study may be based on a particular community which is followed (usually prospectively) over time. For example, a cohort may be based on “all persons aged 20 years or more” living in a particular city or county in a particular year. This would usually require a special survey to be conducted at the start of the follow-up period, with further surveys being conducted at regular intervals.
More specific cohorts
Cohorts may also be constructed not only on the basis of more specific exposures. Perhaps the most common example of this approach involves studies that are based on workers in a particular factory or industry (Checkoway et al, 2004). Such studies may be based on historical records, enabling follow-up to be conducted retrospectively. Typically, such a historical cohort study might involve “all workers who worked for at least one month in the factory at any time during 1970-1999”. The list of such workers can be enumerated using personnel records which also provide information on their job titles and departments (which can be used to estimate their historical exposures).
Comparison populations
In community-based cohorts, comparisons are usually made internally between study participants exposed and those not exposed to a particular risk factor (e.g. low dietary beta carotene intake compared with high dietary beta carotene intake).
In studies of specific populations, an internal comparison may still be possible, e.g. by comparing workers with high benzene exposure to those with low benzene exposure. However, in some instances this may not be possible because good individual exposure information is not available (apart from the fact that workers in the factory received high exposure on the average) or because there is not sufficient variation in exposure within the population (e.g. because everyone who worked in the factory had high exposure). In this situation, an external comparison may be made, e.g. with national death rates or cancer registration rates. In this situation, the source population for the study is effectively the national population, and a comparison is being made between the subgroup in the source population that worked in a particular factory (for example) and the entire source population. Ideally the comparison should be made between the exposed group and the source population minus the exposed group (i.e. everyone else in the country who did not work in the factory). However, this is rarely feasible in practice, and is usually a trivial problem if the exposure is rare. Thus, the comparison is usually made between the exposed group and the national population as a whole.
The risk period
Once the source population has been defined, then the risk period must also be specified. It is important that the risk period is the same for the two or more groups being compared. For example, it would be inappropriate to compare deaths from ischaemic heart disease in two different communities at two different time periods, since there is a continuing decline in IHD mortality, and spurious differences between the communities may be observed if they are not studied over the same risk period.
In a historical cohort study, participants may be followed from some date in the past (e.g. the date the factory opened) up until the present (or some recent date for which death records or cancer registration records are complete). In a prospective cohort study, participants may be followed from the present until some specified future date (e.g. a ten-year follow-up of participants in a recent survey). In both instances, not all study participants will be followed for the entire risk period. For example, someone who moved into the community during the risk period and was “recruited” during a later survey would only be followed from the time of that survey. Similarly, someone who emigrated during the risk period would only be followed until their date of emigration.

Example
The Renfrew/Paisley study was based on two adjacent urban burghs considered to be typical of the West of Scotland. During 1972-1976, men and women aged between 45 and 64 and identified by door-to-door census as living in Renfrew and Paisley were invited to take part. The response rate was 80% (7,052 men and 8,354 women). Participants completed a questionnaire which included self-reported smoking history, occupation, address, age, gender, and respiratory symptoms. Study participants were “flagged” at the National Health Service Central Register in Edinburgh and followed for 20 years. Hart et al (2001) reported that high lung cancer mortality risks were seen for manual compared with non-manual workers. The risk reduced when adjusted for smoking, and reduced further when adjusted for lung function, phlegm and (area) deprivation category. They concluded that the social class difference in lung cancer mortality was explained by poor lung health, deprivation and poor socio-economic conditions throughout life, in addition to smoking.
Example
Rafnsson et al (2001) studied cancer incidence in a cohort of 1690 flight attendants working with two airline companies in Iceland. The total number of person-years of follow-up was 27,148. Among the 1,532 women flight attendants, there were 64 cases of cancer, whereas 51.6 were expected on the basis of national cancer incidence rates (RR=1.2). There was a particularly elevated risk for breast cancer in those who had been hired in 1971 or later and therefore had had the heaviest exposure to cosmic radiation at a young age (RR=4.1). The authors concluded that the association may be due to cosmic radiation or disturbance of circadian rhythm.

Measuring exposure

there are a variety of possible methods for measuring exposure in cohort studies. These include routine records, questionnaires, environmental measurements, Job-Exposure-Matrices (JEM), quantified personal measurements, and biomarkers of exposure.
Ideally, exposures should be measured continuously, or at least at regular intervals, through the risk period (i.e. the period of follow-up). For some risk factors (e.g. for demographic factors such as age, gender and ethnicity), the risk factor status is unlikely to change during the risk period, and can simply be ascertained at baseline. For other exposures that do change over time (e.g. smoking, diet, occupational exposures) regular surveys, or regular examination of routine records, may be desirable to update the exposure information. However, in many studies this is not feasible and information is only collected in a baseline survey; it is then necessary to assume that the exposure level (e.g. serum cholesterol level) has not changed meaningfully during the subsequent follow-up.
In occupational studies, more detailed exposure information may be available through the combination of personnel records (which include changes of job title and department) and Job-Exposure-Matrices (JEMs) based on workplace exposure surveys and/or personal measurements in a subgroup of the workforce
Example
Prescott et al (2004) studied vital exhaustion (fatigue, hopelessness and depression) as a risk factor for ischaemic heart disease (IHD) in 4084 men and 5479 women in Copenhagen. The study was based on participants in the Copenhagen City Heart Study, and the analyses were based on 10,135 people who attended the third follow-up examination in 1991-1993. Cardiovascular risk factors were assessed by a self-administered questionnaire checked with the participant by trained staff, and by various laboratory tests. Vital exhaustion was assessed using a 17-item questionnaire. Participants were followed until 31 December 1997 for fatal and non-fatal IHD, with the information being obtained from the National Board of Health and National Hospital Discharge Register respectively. Subjects with self-reported and verified IHD prior to enrolment were excluded. During follow-up, 483 experienced an IHD event, of which 25% were fatal, and 1559 subjects died from all causes. All but 4 of the 17 items were significantly associated with IHD with significant relative risks ranging from 1.36 to 2.10. The RR for IHD in those with a vital exhaustion score of 10 or more was 2.57 (95% CI 1.65-4.00) and this altered little after adjustment for biological, behavioural and socioeconomic risk factors.

Follow-up
Vital status ascertainment
In some instances, particularly in community based studies, follow-up may involve regular contact with the study participants, including repeated surveys of health status. Perhaps more commonly, follow-up may not involve further contact with the study participants, but may be done by routine record linkage.
For example, study participants may be followed over time by linking the study information with national death records, or incidence records (e.g. a national cancer registry) as well as with other record systems (e.g. social security records, drivers license records) to confirm vital status in those who are not found to have died during the follow-up period.
Although most developed countries have complete systems of death registration, and it is easy in theory to identify all deaths in a particular cohort, this may not be so straightforward in practice. For example, many countries do not have national identification numbers and record linkage may have to be done on the basis of name and date of birth. This may not be infallible because of differences in spelling of names, or inaccuracies in date of birth, but various record linkage programmes are available to identify “near matches” (Jones and Sujansky, 2004). These will be ineffective, however, for people who have changed their name, e.g. because of marriage.
A further problem is that some countries do not have national death registrations, and these may be done on a regional or state basis instead, making it necessary to search multiple registers. Since 1979 a National Death Index for the United States has been compiled and computerized and is available for vital status tracing (Wentworth et al, 1983).
Just because someone has been not been identified in death records, this does not mean that they are still alive and “at risk” since they may have emigrated or may not have been identified in death registrations for some other reason. It is therefore desirable to confirm that they are alive using other record sources such as drivers license records, voter registrations, social security records, etc. In the United States, the Social Security Administration (SSA) records have been frequently used in the past, and in Great Britain the Central Record Office of the Ministry of Pensions and National Insurance is the analogous tracing source (Checkoway et al, 2004).
Coding of the disease outcome
It is not only necessary to determine if and when an event such as a death or hospital admission occurred. It is also necessary to verify, for example, the cause of death, or the cause of a hospital admission. Coding of causes of death should be performed by a nosologist trained in the rules specified by the International Classification of Diseases (ICD) volumes compiled by the World Health Organisation. In many countries this is done routinely for national death registration records, and it is not necessary (or desirable) to recode death registrations for a specific study. However, the ICD codes have changed over time, and when using routine death registration records it is necessary to be aware of which ICD revision was in effect at the time of death.
Person-time
In a study of a specific population, e.g. workers in a particular factory, participants may enter the study on the date that the study starts (1/1/70), or the date that they first meet the eligibility criteria (i.e. employment for one month), whichever is the latest date. If they started working in the factory after the start of the study, then they would only start being followed on the date they started work (or a subsequent date when they met the eligibility criteria).
They stop contributing person-time when they die (or are diagnosed with the disease in an incidence study), emigrate, they are lost to follow-up, or the study finishes (31/12/99) whichever is the earliest.

Example 9.4
Munk Nielsen et al (2003) studied long-term mortality after poliomyelitis by identifying a group of 5,977 patients diagnosed with poliomyelitis in Copenhagen between 1919 and 1954. This involved a review of more than 80,000 consecutive hospital records for Blegdamshospitalet which served as the primary centre for diagnosing and treating patients with acute poliomyelitis in the area of greater Copenhagen. Information extracted from the records included name, sex, date and place of birth, date of admission and discharge, and details of the acute severity of the case.
Since 1 April 1968, all Danish citizens have been given a unique identification number, which is recorded in the Danish Civil Registration System (CRS). The cohort was linked to the CRS to identity individual CRS numbers which were then used to identify deaths in the Danish Cause-of-Death Register. Patients not identified in the CRS were believed to have died or emigrated before 1 April 1968 and for these patients the Cause-of-Death Register was searched for their name and date of birth.
Patients were followed from the initiation of the Cause-of-Death Register in 1943 or the month after the hospital discharge (whichever came later) until the date of death, emigration or 1 May 1997 (whichever came earlier).
There were 1295 deaths compared with an expected number of 1141 (SMR 1.14, 95% CI 1.07-1.20). Excess mortality was restricted to polio patients with a history of severe paralysis of the extremities (SMR = 1.69, 95% CI 1.32-2.15) or patients who had been treated for respiratory failure during the epidemics (SMR = 2.71; 95% CI 2.18-3.37).

Summary
Cohort studies provide the most comprehensive approach for evaluating patterns of exposure and disease, since they involve studying the entire source population (assuming that there is a 100% response rate) over the entire risk period.
Thus, the cohort design ideally includes all of the relevant person-time experience of the source population over the risk period. A cohort study may be based on a particular community (e.g. a geographical community), or on a more specific population defined by a particular exposure (e.g. workers in a particular factory). In both instances, an internal comparison would ideally be made between those participants exposed and those participants not exposed to a particular risk factor. However, in some instances, all of the study participants may be exposed, or valid individual exposure information may not be available, and it may be necessary to make an external comparison, e.g. with national mortality rates (in which case the national population comprises the source population for the study). It is important that any comparisons are made over the same risk period, and that follow-up is as complete as possible. The basic effect measures in a cohort study are the rate ratio and risk ratio.