Anmeldung zur Tagung
Registration
Grußwort
Invitation
Themen
Topics
Allgemeiner Ablauf
General overview
Wissenschaftliches Programm
Scientific program
Kurse
Courses
Symposien
Symposiums
Frühstück mit Spezialisten
Breakfast with specialists
Arzthelferinnen-Fortbildung
Rahmenprogramm
Social program
DOG Information
DOG Information
Allgemeine Informationen
General Information
Autorenindex
Index of Authors
Ausstellerliste
Exhibitors
Sponsoren
Sponsors
Teilnahmegebühren
Registration fees
Impressum
Characteristics of epidemiological studies in ophthalmology
Krummenauer F., Kersten I.
Koordinierungszentrum für Klinische Studien der Universität Mainz, Klinik für Augenheilkunde der Universität Mainz, Germany
Objective: The generalisability of an epidemiological trial crucially
depends on its design, in praticular when phase IV trial in ophthalmology
are considered, where the specific problem of "paired eye data"
has to be taken into account during both planning and analysing the trial.
Therefore the three most important study designs in epidemiology are reviewed
with regard to their main advantages and limitations concerning feasibility
in performance and interpretation of results.
Methods: Based on the description of the planning of an epidemiological
trial to estimate the prophylactic relevance of intra-ocular intra-operative
antibiotics as a protective agent against endophthalmitis after cataract
surgery, the designs of cohort, case control and of cross sectional trials
are opposed. In particular it will be illustrated, which sources of bias
can arise from these designs and therefore could confound the study results.
Results: The most representative and generalisable results can
be obtained from (prospective) cohort studies, since the proof of a causal
relevance of the prophylaxis becomes possible and the study results are
not biased by subject selection. The main advantages of case control trials
are their high efficiency in performance concerning time and costs as
compared to cohort trials; further they allow for sensitive assessment
of confounders to be taken account for during statistical analysis. Their
man drawback refers to possibly biased results due to subject selection
and the therefore lacking chance to prove clinical causality. The cross
sectional design is appreciated for its fast implementation and high cost
effectiveness, but can only provide a static impression of populations
and therefore not estimate the longitudinal impact of intervention programs.
Conclusion: The (prospective) cohort study has to be regarded as
the design of primary choice in the implementation of epidemiological
trials, despite its usually rather large sample sizes and costs.