Implications of different methods for literature searching and assessment in systematic reviews and meta-analyses: A case study

ID: 

2085

Session: 

Poster session 2 Thursday: Evidence synthesis - methods / improving conduct and reporting

Date: 

Thursday 14 September 2017 - 12:30 to 14:00

Location: 

All authors in correct order:

Vinceti M1, Filippini T1, Cilloni S1, Platt J2, Jess C2
1 Research Center for Environmental, Genetic and Nutritional Epidemiology, University of Modena and Reggio Emilia, Italy
2 Cochrane Gynaecological, Neuro-oncology and Orphan Cancers, United Kingdom
Presenting author and contact person

Presenting author:

Marco Vinceti

Contact person:

Abstract text
Background: The development of an adequate literature search strategy when performing a systematic review and meta-analysis of epidemiologic studies may represent a key methodological issue.

Objective: We explored how adding citation chasing to a standard literature search may modify summary estimates computed in a systematic review and meta-analysis.

Methods: We searched the literature through online databases about the effect of supplemental potassium intake on blood pressure in hypertensive individuals, from the early available date till February 2016.

Results: We retrieved 316 records in MEDLINE, 335 in Embase and 309 in CENTRAL. After de-duplication and title/abstract screening, 26 eligible studies were identified and 22 could be included in the meta-analysis. A further extended search based on backward and forward citations of relevant articles and other resources, and particularly on citation chasing, allowed us to identify 7 additional studies, 3 of which eventually eligible for the meta-analysis.

Using the conventional search, overall potassium supplementation was found to decrease systolic blood pressure (SBP) of 3.64 mmHg (95% confidence interval (CI)2.12 to 5.15) and diastolic blood pressure (DBP) of 2.13 mmHg (95% CI 0.48 to 3.79). Using the extended search, SBP decreased by 4.48 mmHg (05% CI 3.07 to 5.90) and DBP by 2.96 mmHg (95% CI 1.10 to 4.82). Little difference between the studies retrieved by the two methodologies emerged by assessing the quality of evidence using the GRADE approach and the risk of bias with the RoB 2.0 tool.

Conclusion: Traditional literature search strategies generally retrieve most of the relevant studies for systematic reviews and meta-analyses. However, they may fail to identify all relevant studies potentially eligible for the analysis, thus affecting to some extent the validity of the summary estimates. Unconventional strategies using backward and forward reference searching may substantially improve the completeness of the literature.