Introduction | Part 1: Hierarchy of Evidence | Part 2: Case Reports | Part 3: Cross-Sectional Studies | Part 4: Case Control Studies | Part 5: Cohort Studies | Part 6: Randomized Controlled Studies | Part 7: Meta-Analysis | Part 8: Systematic Reviews
Cohort studies are observational studies to determine the etiology of a disease based on the presence or absence of a variable. Inherent to their observational nature, they are prospective. This means they start by identifying patients for assignment to study and control groups prior to knowing which ones will develop the disease. The term cohort study applies here in the epidemiological sense; a group defined by demographics, exposure to an intervention, etc.
Again, the case-control study is a retrospective study which identifies patients with the disease first and then looks backward looking for potential causation. The cohort study is prospective by allocating target and control groups first and then following patients through to identify which ones develop the target disease and those that don't.
Case-Control (retrospective - time flows backward):
Potential causes ← Retrospective review ← CASES
Potential causes ← Retrospective review ← CONTROLS
Cohort (prospective - time flows forward):
Exposed (study cohort) → Follow-up → Incidence of outcome
Not Exposed (control cohort) → Follow-up → Incidence of outcome
Both study designs are observational. Target characteristics are reviewed through follow-up (cohort) or by retrospective review (case-control). In contrast to the clinical trial, there is no intervention designed to alter an outcome. Investigators do not predetermine who will receive the target exposure or intervention. Simply, those who have been exposed and those who have not are chosen for longitudinal observation.
|A control group is not an essential component of a cohort study and some studies do not include one. When the objective is to assess a cause and effect relationship, the cohort study is superior to case-control studies because of greater confidence that the influence (cause) preceded the disease (effect). However, cohort studies require a larger sample size when compared to case-control studies to reach the same statistical potential and to follow patients over time entails greater costs.|
Concurrent cohort studies employ the basic design of this type and follow patients over a considerable length of time. To reduce study time and cost, a database study (often called a non-concurrent cohort study or retrospective or historic or database study), assignment of subjects are made from reliable data obtained on the patient (e.g. health record databases). Existing records are used to define a cohort with known past exposures and then these patients are then observed forward in time or may simply be followed to the present. To avoid assignment bias, the researcher must not know the features of the patients being evaluated until after they are assigned to their respective groups.
Cohort studies may help to further validate proposed causal relationships in retrospective studies. There is an explicit hypothesis which can be better tested with controls. However, as previously mentioned, their prospective nature involves considerable time and cost to the study. The longer the study, the greater chance of patients who 'drop-out' for various reasons (e.g. death). This issue of 'latency' should be addressed when comparing cohort studies with similar populations and exposure, but with significant differences in the period of observation. Because of the relatively limited number of subjects, cohort studies are not well suited for rare diseases as well. However, because of the longitudinal observational assessment of exposure inherent to the cohort, it is the predominant method used in the study of cancer.
The position of the cohort study below the clinical trial in the evidence hierarchy is largely because there is a greater risk of systematic bias. Co-variables such as risk factors may affect assignment of patients to a specific group and this bias may negatively effect the outcome. Confounding variables, features which may not appear to directly affect the outcome, but may have a potentially significant effect on the observed outcome, need to be carefully evaluated. For example, a hypothetical study assessing the relationship between obesity and diabetes in different breeds of cats may find that purebreds are at higher risk of developing diabetes. However, a potentially ignored confounder is that there is greater likelihood of obesity in purebred cats because purebred cat owners may provide a different environment featuring a more sedentary lifestyle and overfeeding. Mixed breed cats perhaps have a higher chance of experiencing a relatively feral, outdoor lifestyle with greater exercise, etc. Therefore the confounding variable of breed in relation to lifestyle, would systematically bias results in favor of purebred cats having a higher risk of diabetes.
From 2000-2010, case-control studies were reported most often in JAVMA (1,975), Veterinary Surgery (600), Journal of Veterinary Internal Medicine (536), Equine Veterinary Journal (307), Journal of the American Animal Hospital Association (301), Veterinary Record (293), Journal of Small Animal Practice (278), Australian Veterinary Journal (223), Veterinary Pathology (184) and Veterinary Radiology and Ultrasound (176).
Top concept representation (MeSH Terms), excluding cohort studies were retrospective studies, dogs and dog diseases, prospective studies, horses and horse diseases, treatment outcome, cats and cat diseases and cattle, respectively.
Association between results of ambulatory electrocardiography and development of cardiomyopathy during long-term follow-up of Doberman pinschers. Calvert CA, et al, J Am Vet Med Assoc 2000 Jan 1;216(1):34-9
In this study 114 Doberman Pinscher dogs (a breed genetically predisposed to dilated cardiomyopathy - DCM), who were assessed to be overtly healthy at the start of the study (no echocardiographic abnormalities), were followed for at least 2 years and as long as 10 years from the start of the study. At predetermined intervals, ambulatory electrocardiography (24 hours) and echocardiography were obtained. The outcome showed that the development of ventricular premature contractions showed an increased likelihood with both age and with males. A variety of outcomes including sudden onset death, congestive heart failure and death due to noncardiac causes (n=2) were observed. The conclusions at the end of this long-term study was that the results of high-quality ambulatory ECG (Holter) recordings are useful in identifying overtly healthy Doberman Pinschers who are at high risk for developing DCM.
This series has been loosely organized from a set of lectures given by the author within graduate courses in Biomedical Informatics beginning in 2002. Content is being edited to improve organization, depth, correct inaccuracies as well as updates with new information during Winter/Spring 2012. Feedback is greatly appreciated. © 2011, Stuart Turner.