Epidemiological and reporting characteristics of systematic review updates

Article type
Authors
Rombey T1, Lochner V1, Puljak L2, Koensgen N1, Mathes T1, Pieper D1
1Institute for Research in Operative Medicine, Witten/Herdecke University
2Center for Evidence-based Medicine and Health Care, Catholic University of Croatia
Abstract
Background: while there is guidance on when and how to update systematic reviews (SRs), to date there are no clear reporting guidelines for SR updates. Guidance on SRs, such as the PRISMA statement, is not entirely relevant to SR updates, which have their own specific aspects that need to be reported. The lack of reporting guidelines poses problems for authors wanting to publish an SR update and means that existing SR updates are probably reported inconsistently with varying levels of informative value.

Objectives: to investigate epidemiological and reporting characteristics of recently published SR updates and thereby inform the development of reporting guidelines for SR updates.

Methods: we searched PubMed for SR updates published between 1 January 2016 and 22 January 2018. We excluded updates of individual patient data or network meta-analyses, clinical guidelines, Cochrane Reviews or SRs with more than one key question. We included a randomly selected sample of n = 100 updates of SRs of interventions, whose full texts were available in English or German. We extracted and analyzed data on various epidemiological, descriptive and reporting characteristics using descriptive statistics.

Results: most SR updates analyzed non-pharmacological interventions (64/100). They were published a median of five years (interquartile range (IQR) 4 to 8) following publication of the previous SR’s edition and included a median of 20 (IQR 9 to 28) studies and 2028 (IQR 1041 to 5244) participants. A median of seven (IQR 4 to 16) studies comprising a median of 1501 (IQR 329 to 4037) participants had been newly identified since the previous edition of the SR. Nineteen SR updates did not include any new study compared to the previous version.

In 78 SR updates, less than 50% of the authors remained the same. Only in half the SR updates (52/100), the word ‘update’ was part of the title. No SR update reported using a decision framework or prediction tool to justify the decision to update. Thirty-one SR updates, reported a change of the PICOs (patient, intervention, comparison, outcome) and 42 SR updates a change of at least one methodological aspect (study selection process, n = 26; search strategy, n = 23; statistical analysis, n = 11; data extraction process, n = 1). The number of studies and participants included in the previous edition was only reported in 66 and 26 SR updates, respectively. No SR update used tables or graphs to compare the results of the update with the former edition. A change in the conclusions was reported in 31 SR updates.

Conclusions: future reporting guidelines for SR updates should focus particularly on appropriate labelling of and justification for SR updates. Furthermore, they should provide details of how information related to the previous edition, such as number of included studies and participants, and results, should be incorporated in the SR update.

Patient or healthcare consumer involvement: since this was a methodological study, there was no patient or healthcare consumer involvement.