Biostatistics Types of Studies Research classifications Observational vs

Biostatistics ~ Types of Studies

Research classifications Observational vs. Experimental Observational – researcher collects info on attributes or measurements of interest, but does not influence results. Experimental – researcher deliberately influences events and investigates the effects of the intervention, e. g. clinical trials and laboratory experiments.

Research classifications Prospective vs. Retrospective Prospective – data are collected forwards in time from the start of the study, e. g. experiments & survival studies. Retrospective – data refer to past events and may be acquired from existing sources such as hospital records or by interview, e. g. case-control studies & cohort studies.

Research classifications Longitudinal and Cross-Sectional Longitudinal – investigates changes over time, e. g. survival studies. Cross-sectional – individuals are observed only once, e. g. surveys.

Example 1 – Cholesterol Drug Test (i. e. Phase III clinical trial) How might we proceed?

Example 2 – Determine Whether Age at 1 st Pregnancy is a Risk Factor for Cervical Cancer How might we proceed?

Example 3 – Design an experiment to compare two treatments of ovarian cancer patients. How might we proceed? What are some potentially unique issues/challenges for such a study?

Example 4 – Determine if right heart catheterization (Swan-Ganz line) increases risk of 30 -day mortality amongst heart patients. How might we proceed?

Selection of Controls in Case. Control Study Goal: Want controls as similar as possible to cases. What is typically done: § Use patients hospitalized for other reasons § Matching according to some criteria. Only helpful if variables used are strongly related to both the risk factor and the outcome of interest. § Usually have more controls than cases, especially with rare events. Can have multiple matched controls per case.

Selection of Cases Generally must meet some entrance criteria (but may have to take what one can get) Need to take into account or realize that all cases are not similar in cause disease, exposure to risk factor, degree of disease, etc…

Bias in Case-Control Studies A diseased person is likely to recall more risk factors (especially widely publicized ones) than a healthy person. This is called perceived risk. Missing data in medical records Presence of risk factor may increase chance of detection In accuracy in reporting of exposure, e. g. How much did you drink? How much did you smoke? Etc…

Cohort Studies Method of choice for observational studies The essence of the cohort study is identify a group of subjects and then follow them up to see what happens (longitudinal). Compare to age at 1 st pregnancy and cervical cancer study. Quality of data recording can be carefully controlled.

Problems with Cohort Studies Selection of subjects (lots of issues) Loss to follow-up (biggest problem) Surveillance Bias – high risk group might be studied more carefully. Best approach is to investigate all subjects identically and assessors should be “blinded”.

Cross-Sectional Study All information is collected at the same time. Sampling is always an issue Nonresponse bias Volunteer bias As with all observational studies it is hard if not impossible to establish causation!