Full Text Screening

Full text screening is the stage in a systematic review where the complete texts of potentially relevant studies, identified during title and abstract screening, are retrieved and read in full. Each study is assessed against the predefined eligibility criteria, and studies that do not meet these criteria are excluded with documented reasons.^[1][2] This stage aims to ensure that only truly eligible studies enter the data extraction and synthesis phases.

Title and abstract screening is necessarily conservative: reviewers include any study that might be eligible based on limited information. Full text screening resolves the uncertainty by providing complete information about the study's population, intervention, comparator, outcomes, study design, and other eligibility parameters. It is the definitive eligibility decision. The quality of full text screening directly affects the validity of the systematic review; including ineligible studies introduces noise or bias, while excluding eligible studies reduces the comprehensiveness of the evidence base. Transparent documentation of exclusion reasons at this stage (which PRISMA 2020 asks authors to report, in item 16b) allows readers to evaluate whether the review team applied criteria consistently and reasonably.^[2][3]

After title and abstract screening, the review team retrieves the full text articles of all studies that passed or were marked as uncertain. Full texts are typically obtained from library databases, interlibrary loan, or by contacting study authors.

Two reviewers independently assess each full text against the predefined eligibility criteria. For each excluded study, the reviewer records the primary reason for exclusion (for example, "wrong population," "wrong study design," "wrong outcome," or "wrong comparator"). When the two reviewers disagree, they discuss the discrepancy and, if needed, a third reviewer resolves the disagreement.^[1]

The results of full text screening are reported in the PRISMA 2020 flow diagram, which shows the number of reports sought for retrieval, the number that could not be retrieved, the number assessed for eligibility, the number excluded (with reasons), and the numbers of studies, and of reports of those studies, included in the review.^[2] A table or appendix listing each excluded study and its exclusion reason is often included in systematic review publications or supplementary materials.

Several practical challenges arise during full text screening. Some studies may have multiple publications (duplicate reporting), requiring careful linking so that each study, rather than each report, is the unit of the review and the same data are not counted twice.^[1] Some full texts may be in languages the reviewers do not read, requiring translation or assistance from colleagues. Occasionally, full texts are unavailable despite reasonable efforts, which is documented as well.^[2]

Full text screening is a standard step in systematic reviews and is conducted after title and abstract screening and before data extraction. It applies to all types of systematic reviews, including reviews of interventions, diagnostic tests, prognostic studies, qualitative evidence, and economic evaluations. The step cannot be skipped; title and abstract information alone is insufficient for making definitive inclusion decisions, although clearly irrelevant records can be excluded at the title and abstract stage.

A systematic review of exercise interventions for chronic low back pain identifies 342 potentially relevant reports after title and abstract screening. The team seeks all 342 full texts and obtains 334; 8 cannot be retrieved despite contacting the authors. Two reviewers independently screen each of the 334 against the eligibility criteria and exclude 233: wrong study design (92), wrong population (68), wrong intervention (45), and wrong outcome (28). The remaining 101 reports describe 87 studies, because 14 are additional publications of studies already included; these are linked to their study rather than excluded or counted twice. The PRISMA 2020 flow diagram documents these numbers (342 sought, 8 not retrieved, 334 assessed, 233 excluded with reasons, 87 studies from 101 reports included), and an appendix lists each excluded report with its reason.