Biases

1) SYSTEMATIC ERROR:

In an analytical study, this may be produced in the induction of subjects in the study, in their allocation to the treatment groups and in the collection, analysis, interpretation, publication and review of the data.

2) BERKSON’S BIAS:

Type of selection bias, which may occur in case-control studies, in which the fact of being ill and having been exposed to the risk factor under study increases the probability of being admitted to hospital, which gives rise to a systematically higher rate of exposure among hospital cases, in comparison with the control cases, also in hospital.

3) INTERVIEWER BIAS:

Systematic error, variant of information bias, due to the selective collection of data, consciously or unconsciously, by the interviewer.

In order to reduce this, structured questionnaires are used.

4) INFORMATION BIAS:

Systematic error resulting from measuring the exposure (case-control studies) or the evolution (cohort studies and non-blinded clinical trials) with different intensity between the two groups compared.

5) MEMORY BIAS:

Systematic error due to differences when recalling previous facts or experiences.

In case-control studies, the cases may recall some previous experiences more than the controls, which in principle leads to an overestimation of the odds ratio (ORA).

The bias due to memory sensitization is a variant of memory bias, in which the memory of the two groups of patients has not been sensitized in the same way; the structured questionnaire interview helps to reduce it.

6) OBSERVATION BIAS:

Systematic difference between the real and recorded value.

In a clinical trial, the most common one is that due to knowledge, by the observer, of the treatment received by each subject.

7) PUBLICATION BIAS:

Type of selection bias, which is the result of thinking that published trials (and other studies) are the ones actually carried out.

Many clinical trials are not published (because they aren’t finished, because the researcher considers the results irrelevant, because the sponsor doesn’t want to publish them or because they are not accepted for publication).

This occurs especially with clinical trials which do not register differences between the different treatment groups.

Other trials are published in more than one medical journal (duplicated publication), on occasions signed by different authors, making it difficult to identify some duplicated publications.

Both phenomena tend to determine that the meta-analyzes which include only the results of published clinical trials in general overestimate the effect of the experimental treatment.

8) SELECTION BIAS:

Systematic error as a result of the fact that the characteristics of the subjects included in a study are different from the characteristics of those not included, which means that the sample is not representative of the reference population.

9) TYPE-I ERROR:

Error which consists of rejecting a null hypothesis, which is true.

The alpha error is the error made when a null hypothesis is rejected when it is true.

The alpha probability is the probability of making a Type-I error. It is usually fixed at a = 0.05, that is, a probability of 5%.

10) TYPE-II ERROR:

Error which consists of accepting a null hypothesis, which is false.

The beta error is the error made when a null hypothesis is not rejected when it is false.

The beta probability is the probability of making a Type-II error.

It is usually fixed at 10% (so that 1- β = 0.90) or at 20% (so that 1- β = 0.80). The term 1-β is the statistical power of the study.

11) TYPE-III ERROR:

Error which consists of considering a treatment as superior when it is really inferior.

12) GLOBAL ALPHA ERROR:

It is the alpha error which is made by making multiple comparisons.