RLO: Steps in conducting a systematic review

Sections

Step Three

Step 3 is about assessing the quality of the included studies. Quality refers to the internal validity. The internal validity of a study refers to the degree which its results are likely to be free from bias.

Bias

'Bias is any process at any stage of inference tending to produce results that differ systematically from the true values'.

There are four main sources of systematic bias in trials of the effects of healthcare.

Selection bias - Resulting from the way comparison groups are assembled.
Performance bias - Systematic differences in care given to the participants.
Detection bias - Systematic differences between the comparison groups in outcome assessment.
Attrition bias - Systematic differences between the comparison groups in the loss of participants in the study.

Selection bias

The idea of randomisation to generate groups that are comparable in terms of known or unknown potential confounding factors. Randomisation consist of two parts:

generation of allocation sequences
concealment of allocation sequences

Examples of adequate generation of allocation sequences, in other words, where the resulting sequences are unpredictable are computer generated random numbers, drawing lots, tossing a coin or throwing a die.

Inadequate generation of allocation sequences would include: organising cases by case number, date of birth, date of admission or alternation.

Examples of adequate concealment of allocation sequences, this is where the enrolling investigators cannot foresee assignment are: a priory of numbered or coded drug containers prepared by an independent pharmacy, central randomisation performed at a site away from the trial location, sequentially numbered, sealed , opaque envelopes.

Inadequate conealment of allocation sequence where the enrolling investigators can foresee upcoming assignment are: all procedures based on inadequate generation of allocation assignment, open allocation schedule, unsealed or non-opaque envelopes.

Concealment has been shown to be more important in preventing bias than other components of allocation, such as generation of the allocation sequence.

Performance bias

The refers to the systematic differences in the care given to the participants in the comparison groups, other than the intervention under investigation. Therefore to protect against unintended differences in care and placebo effects, those providing and receiving care may be 'blinded' so they are unaware of the group to which the receiver of care has been allocated. There is good evidence that those who are aware of their assignment status report more symptoms, which may lead to biased results.

Detection bias

Occurs if the knowledge of patient assignment influences the process of outcome assessment. In order to get around this, outcome assessors and patients may be blinded.

Attrition bias

Attrition bias may be introduced when protocol deviations and loss to follow-up leads to the exclusion of patients after they have been allocated to treatment groups.