RESEARCH SERIES Observational research methods. …

1 RESEARCH SERIESO bservational RESEARCH methods . RESEARCH design II:cohort, cross sectional, and case-control studiesC J Med J2003;20:54 60 Cohort, cross sectional, and case-control studies arecollectively referred to as Observational studies. Oftenthese studies are the only practicable method ofstudying various problems, for example, studies ofaetiology, instances where a randomised controlled trialmight be unethical, or if the condition to be studied israre. Cohort studies are used to study incidence, causes,and prognosis. Because they measure events inchronological order they can be used to distinguishbetween cause and effect. Cross sectional studies areused to determine prevalence. They are relatively quickand easy but do not permit distinction between causeand effect. Case controlled studies compare groupsretrospectively. They seek to identify possible predictorsof outcome and are useful for studying rare diseases oroutcomes. They are often used to generate hypothesesthat can then be studied via prospective cohort or , cross sectional, and case-controlstudies are often referred to as observa-tional studies because the investigator sim-ply observes.

2 No interventions are carried out bythe investigator. With the recent emphasis on evi-dence based medicine and the formation of theCochrane Database of randomised controlledtrials, such studies have been somewhat gliblymaligned. However, they remain important be-cause many questions can be efficiently answeredby these methods and sometimes they are theonly methods objective of most clinical studies is todetermine one of the following prevalence, inci-dence, cause, prognosis, or effect of treatment; itis therefore useful to remember which type ofstudy is most commonly associated with eachobjective (table 1)While an appropriate choice of study design isvital, it is not sufficient. The hallmark of goodresearch is the rigor with which it is conducted. Achecklist of the key points in any study irrespec-tive of the basic design is given in box published study should contain suffi-cient information to allow the reader to analysethe data with reference to these key this article each of the three importantobservational RESEARCH methods will be discussedwith emphasis on their strengths and weak-nesses.

3 In so doing it should become apparentwhy a given study used a particular researchmethod and which method might best answer aparticular clinical STUDIEST hese are the best method for determining theincidence and natural history of a condition. Thestudies may be prospective or retrospective andsometimes two cohorts are cohort studiesA group of people is chosen who do not have theoutcome of interest (for example, myocardial inf-arction). The investigator then measures a varietyof variables that might be relevant to the develop-ment of the condition. Over a period of time thepeople in the sample are observed to see whetherthey develop the outcome of interest (that is,myocardial infarction).In single cohort studies those people who donot develop the outcome of interest are used asinternal two cohorts are used, one group hasbeen exposed to or treated with the agent ofinterest and the other has not, thereby acting asan external cohort studiesThese use data already collected for otherpurposes.

4 The methodology is the same but thestudy is performed posthoc. The cohort is followed up retrospectively. The study periodmay be many years but the time to complete thestudy is only as long as it takes to collate and ana-lyse the and disadvantagesThe use of cohorts is often mandatory as arandomised controlled trial may be unethical; forexample, you cannot deliberately expose people tocigarette smoke or asbestos. Thus RESEARCH on riskfactors relies heavily on cohort cohort studies measure potential causesbefore the outcome has occurred the study candemonstrate that these causes preceded theTable 1 ObjectiveCommon designPrevalenceCross sectionalIncidenceCohortCause (in order ofreliability)Cohort, case-control, crosssectionalPrognosisCohortTreatment effectControlled ,Departmentof Accident and EmergencyMedicine, Taunton andSomerset Hospital,Taunton, Somerset, UKCorrespondence to:Dr C J on July 2, 2022 by guest. Protected by Med J: first published as on 1 January 2003.

5 Downloaded from outcome, thereby avoiding the debate as to which is cause andwhich is further advantage is that a single study can examinevarious outcome variables. For example, cohort studies ofsmokers can simultaneously look at deaths from lung, cardio-vascular, and cerebrovascular disease. This contrasts withcase-control studies as they assess only one outcome variable(that is, whatever outcome the cases have entered the studywith).Cohorts permit calculation of the effect of each variable onthe probability of developing the outcome of interest (relativerisk). However, where a certain outcome is rare then aprospective cohort study is inefficient. For example, studying100 A&E attenders with minor injuries for the outcome ofdiabetes mellitus will probably produce only one patient withthe outcome of interest. The efficiency of a prospective cohortstudy increases as the incidence of any particular outcomeincreases. Thus a study of patients with a diagnosis of deliber-ate self harm in the 12 months after initial presentation wouldbe efficiently studied using a cohort problem with prospective cohort studies is the lossof some subjects to follow up.

6 This can significantly affect theoutcome. Taking incidence analysis as an example (incidence= cases/per period of time), it can be seen that the loss of a fewcases will seriously affect the numerator and hence the calcu-lated incidence. The rarer the condition the more significantthis studies are much cheaper as the data havealready been collected. One advantage of such a study designis the lack of bias because the outcome of current interest wasnot the original reason for the data to be collected. However,because the cohort was originally constructed for anotherpurpose it is unlikely that all the relevant information willhave been rigorously cohorts also suffer the disadvantage thatpeople with the outcome of interest are more likely to remem-ber certain antecedents, or exaggerate or minimise what theynow consider to be risk factors (recall bias).Where two cohorts are compared one will have beenexposed to the agent of interest and one will not. The majordisadvantage is the inability to control for all other factors thatmight differ between the two groups.

7 These factors are knownas confounding confounding variable is independently associated withboth the variable of interest and the outcome of interest. Forexample, lung cancer (outcome) is less common in peoplewith asthma (variable). However, it is unlikely that asthma initself confers any protection against lung cancer. It is moreprobable that the incidence of lung cancer is lower in peoplewith asthma because fewer asthmatics smoke cigarettes (con-founding variable). There are a virtually infinite number ofpotential confounding variables that, however unlikely, couldjust explain the result. In the past this has been used to sug-gest that there is a genetic influence that makes people wantto smoke and also predisposes them to only way to eliminate all possibility of a confoundingvariable is via a prospective randomised controlled study. Inthis type of study each type of exposure is assigned by chanceand so confounding variables should be present in equalnumbers in both , problems can arise as a result of bias.

8 Bias can occurin any RESEARCH and reflects the potential that the samplestudied is not representative of the population it was drawnfrom and/or the population at large. A classic example is usingemployed people, as employment is itself associated with gen-erally better health than unemployed people. Similarly peoplewho respond to questionnaires tend to be fitter and moremotivated than those who do not. People attending A&Edepartments should not be presumed to be representative ofthe population at to run a cohort studyIf the data are readily available then a retrospective design isthe quickest method. If high quality, reliable data are notavailable a prospective study will be first step is the definition of the sample group. Eachsubject must have the potential to develop the outcome ofinterest (that is, circumcised men should not be included in acohort designed to study paraphimosis). Furthermore, thesample population must be representative of the generalpopulation if the study is primarily looking at the incidenceand natural history of the condition (descriptive).

9 If however the aim is to analyse the relation betweenpredictor variables and outcomes (analytical) then the sampleBox 1 Study purposeThe aim of the study should be clearly sample should accurately reflect the population fromwhich it is source of the sample should be sampling method should be described and the samplesize should be criteria and exclusions should be stated and number of patients lost to follow up should be stated andexplanations groupThe control group should be easily source of the controls should be explained are they fromthe same population as the sample?Are the controls matched or randomised to minimise biasand of measurements and outcomesValidity are the measurements used regarded as valid byother investigators?Reproducibility can the results be repeated or is there a rea-son to suspect they may be a one off ?Blinded were the investigators or subjects aware of theirsubject/control allocation?Quality control has the methodology been rigorouslyadhered to?

10 CompletenessCompliance did all patients comply with the study?Drop outs how many failed to complete the study?DeathsMissing data how much are unavailable and why?Distorting influencesExtraneous treatments other interventions that may haveaffected some but not all of the factors Are there other variables that mightinfluence the results?Appropriate analysis Have appropriate statistical tests beenused?ValidityAll studies should be internally valid. That is, the conclusionscan be logically drawn from the results produced by anappropriate methodology. For a study to be regarded as validit must be shown that it has indeed demonstrated what it saysit has. A study that is not internally valid should not bepublished because the findings cannot be question of external validity relates to the value of theresults of the study to other populations that is, the generalis-ability of the results. For example, a study showing that 80%of the Swedish population has blond hair, might be used tomake a sensible prediction of the incidence of blond hair inother Scandinavian countries, but would be invalid if appliedto most other RESEARCH on July 2, 2022 by guest.