Study Designs in Epidemiology

For formation and evaluation of causal hypothesis and for

Comparison of groups of animals

Dr. Bhoj R Singh, Principal Scientist (VM)I/C Epidemiology; Centre for Animal Disease Research and DiagnosisIndian Veterinary Research Institute, Izatnagar-243122, Bareilly, UP,

India. TeleFax  +91-581-2302188

Types• Qualitative

– Formulation of hypothesis about causes (unknown) of diseases

– Source of infection– Medical detectives

• Quantitative– Surveys– Monitoring and surveillance – Modeling– Biological and economic evaluation of disease control

programmes– Armchair epidemiology

Several study designs

• Experimental

• Observational– Cohort– Case Control– Cross-sectional– Case-Crossover

Experimental• Trials

– Systematic study to establish the procedure’s prophylactic or therapeutic effect, improvement in production or amelioration of clinical disease.

• Full control of investigator on selecting the study population

• Most reliable and scientific information in properly designed experiments

Types of trials

• Pharmacological and toxicity trials: On experimental animals or on target species• Initial trials of therapeutic effect and safety: On target species, on small scale, in

controlled environment, usually to compare and find the best drug, dose, route of administration, formulation etc.

• Clinical evaluation of efficacy (Field trials): On large scale in field, under operational conditions, management and environment can affect the results.

– Clinical trial: Subject must be selected randomly. On patients, either (mostly) therapeutic or prevention of sequelae.

• Open trial: Clinical condition of a patients is compared before and after treatment, interpretation is difficult, uncontrolled clinical trial

• Controlled clinical trials: Group getting treatment is compared with one not getting the treatment (control). The control may be concurrent or historical (often criticized)

– Field trial: Mostly on clinically healthy individuals to determine prophylactic efficacy.

• Controls:– Positive control: getting standard treatment for the disease– Negative control: Getting no treatment– Placebo: Inert substance visually similar to treatment. (John Haygarth, 1800, to evaluate electromagnetic

influence of metal rods (perkins) insertion for treatment of diseases )– Concurrent control: At the same time (Pasteur’s trail of Anthrax vaccine, 1922)– Randomized controls (Bradford Hill for Pertusis vaccine and for efficacy of Streptomycin in TB)

• Post authorization surveillance: To monitor adverse effect during use.

• Confirmatory trials: Randomized controlled clinical trial to determine appropriate dose levels of drugs, stringent protocol. Usually conducted at several laboratories and at several sites.

• Exploratory trials: Same as above but less stringent protocol, usually conducted at one laboratory or at a single site.

• Community trials: Experimental unit is an entire community, like fluoridation of public water supply in prevention of dental carries, Mineral blocks in entire area.

• Multicentre trials: These are only method to accruing sufficient number of animals, experimental units within a reasonable period of time. They also allow wide representation of population and increases validity.

• Superiority trials: To detect difference between a treated and a control group.

• Equivalence trials and non-inferiority trials: To demonstrate that effect of each treatment is same. Usually to compare an inexpensive treatment over established expensive one.

• Efficacy: the measure to define outcome of the trial.– %efficacy= (C-T)*100/ C– Suppose to evaluate effect of a drug on ecto-parasites a trail was done then

C= Mean number of ecto-parasites/ No. of animals in control group

T= Mean number of ecto-parasites/ no. of animals in treated group

• Experimental units: The smallest independent unit to which the treatment in randomly allocated. May be an elementary unit (an animal), an aggregate as a pen or herd.

• Experimental population: Population in which trial is conducted, should be a representative of target population. Because difference between the two may result in non-generalizable (externally valid).

• Internal validity: Indicate that the observed difference between T and C group is due to the treatment. Can be obtained by good trial design, and selection of good target population (as for prophylactic trail a population with high risk of developing the disease). By randomization or using an good alternative (allocation to control/ placebo and treatment according to date of entry).

Trial designs• Parallel-group (Standard): Commonly used in

confirmatory trials. EUs are randomized to a single treatment group.

• Cross over: When subjects are exposed to more than one treatment consecutively. Each treatment is chosen randomly. Experimental units act as their own control.

• Sequential: A trial, effect of which depends on results so far obtained.

• Factorial: when two treatments (A and B) are to be evaluated at two dose levels (a and b). It gives rise to ab experimental conditions. All possible combinations are of levels and treatments are taken in to consideration.

• Termination of trials: Trials last as long as it takes to enlist the units and for the last unit to complete the trial. Trial is terminated because of:

– Serious adverse effect evident in a trial

– If specified difference is detected to the predetermined level of significance (particularly in sequential trials).

• Meta-analysis: Statistical analysis of data pooled from several studies to integrate findings with goals:

– To increase statistical power of primary end points.– To resolve uncertainty in case of conflicting observations.– To improve estimates of therapeutic effect and their precision– To answer question not posed at the beginning of individual trial– To give a ‘state of art’ literature review– To facilitate analysis of subgroups when individual experiment has low power.– To guide researchers in planning new trials– To offer rigorous support for generalization of a treatment.– To balance ‘overflow of enthusiasm associated with introduction of new

procedure following single or first beneficial report.

Cross-Sectional

• Relationship of diseases is investigated in a population with hypothesized causal factor.

• X 2 test, 2x2 tables are often used to draw conclusion

• Investigator do not have much control over the selection of the study population as almost whole population in a defined region is included in the study.

Co-Hort

• Groups having exposure or no exposure are compared to find out the causal association of the exposure.

• No control of investigator on study sample however adjustments to age, sex, breed etc. are to be made to make the two groups comparable.

Case-Control

• Groups with or without disease are compared and relative and attributable risks are compared to find out the causal association with predicted or hypothesized causal factor.

• Investigator can not exercise control over selection of a case however a bit on control to find the comparable control.

Case-Crossover

• A variant of the case-control design• used to study the effects of transient exposures

on acute events.• This design samples only cases and compares

each case’s exposure during a time period just before the case-defining event (hazard period) with that subject’s own exposure in other reference periods (control periods).

• Each subject serves as its own control; there is perfect matching on all measured or unmeasured subject characteristics that do not vary over time.

• Unidirectional case-crossover designLess used designed, only unidirectional comparison is possible.

• Symmetric bidirectional case-crossover design

• In this design, reference periods are symmetrically spaced in time, both before and after the hazard period, which minimizes potential time-varying confounding by season or time trends.

• In this study, the hazard period is defined as the day of death, and exposure is modeled as the mean of exposure on the day of death (lag0) and the day before death (lag1).

• Reference days (one or more) are matched to each hazard day. Hazard days that could not be matched to the full complement of reference days are excluded rather than sacrificing the symmetry of the matching.

• Selection bias is possible in case-crossover studies when some days in a time series are unevenly sampled as hazard or reference (control) days.

Surveys• Examination of aggregates of EU (a herd may be

an example of an aggregate).• An epidemiological survey essentially include

determination and counting of a certain or all diseases, death, determinants and attributes of disease, health and production.

• Survey may be on sample or on whole population (census)

• Types– Cross sectional– Longitudinal

• Prospective• Retrospective

• Screening: Specific type of diagnostic survey with aim to diagnose undiagnosed cases so that healthy and sick (infected) can be separated. Screening test not necessarily a diagnostic test.– Mass screening: TB-JD testing in a population– Strategic screening: Screening in the areas

where presence of a disease is suspected– Prescriptive screening: Testing for suclinical

mastitis, pap-smear test for cervical cancer

Monitoring and Surveillance

• Monitoring: – Routine observation on health, productivity and

environmental factors.– Recoding the observations– Transmission of of the observations

• Surveillance:– More intensive form of data monitoring and recording– Usually part of a disease control programme– It include Collation and interpretation of data collected

during monitoring programme.