Article type
Year
Abstract
Background: Understanding and improving the prognosis of a disease or health condition has been identified as a priority in clinical research and practice. The number of systematic reviews (SR) of prognostic studies is increasing and guidance exists for their conduct.
Objectives: The aim of this exploratory research was to investigate the nature and quality of SRs of prognostic studies using a random sample from KSR Evidence (a database including all SRs and meta-analyses in healthcare published since 2015).
Methods: A 2016 random sample of SR publications was generated and screened to identify reviews classified as ‘prognostic’. Data were extracted on review type (according to the PROGRESS framework) and study characteristics (including country and disease type classified according to ICD-10). All studies were appraised using an adapted version of the ROBIS (Risk Of Bias In Systematic Reviews) Tool. ROBIS considers four domains: study eligibility criteria; identification and selection of studies; data collection and study appraisal; and synthesis and findings. From these four domains an overall summary of the risk of bias (ROB) is generated.
Results: From a random sample of 516 SRs, 87 (17%) prognostic reviews were identified. Most were SRs of one or more prognostic factors of disease (79, 91%). Two (2%) were classified as fundamental prognostic research, four (5%) considered the development, validation or impact of prognostic models. None considered the use of prognostic information to tailor treatment decisions and two (2%) covered multiple categories. Prognostic SRs were identified in thirteen disease areas with the majority relating to cancer (26, 30%) followed by mental health (10, 11%) and circulatory diseases (10, 11%). Twenty-two countries were represented in the sample with China contributing the most SRs (21, 24%) followed by the USA (14, 16%) and the UK (11, 13%). Generally, SRs were at high ROB across all four domains with domain 2 (identification and selection of studies) being the weakest (85% at high or unclear ROB). Overall, just seven SRs (8.5%) were at low ROB. The main areas of concern were not reporting a comprehensive search strategy, restrictions on sources of information and an inadequate assessment of the quality of the included studies.
Conclusions: SRs of prognostic studies are increasing in number but according to our random sample, despite extensive guidance available, the majority are at high ROB. The literature is dominated by cancer and most SRs investigated prognostic factors rather than models. We aim to use this research as a starting point for further exploration of trends in prognostic SRs.
Patient or healthcare consumer involvement: Whilst no healthcare consumers were involved, gaining a better understanding of the methodology of prognostic research should ultimately result in better healthcare outcomes for patients.
Objectives: The aim of this exploratory research was to investigate the nature and quality of SRs of prognostic studies using a random sample from KSR Evidence (a database including all SRs and meta-analyses in healthcare published since 2015).
Methods: A 2016 random sample of SR publications was generated and screened to identify reviews classified as ‘prognostic’. Data were extracted on review type (according to the PROGRESS framework) and study characteristics (including country and disease type classified according to ICD-10). All studies were appraised using an adapted version of the ROBIS (Risk Of Bias In Systematic Reviews) Tool. ROBIS considers four domains: study eligibility criteria; identification and selection of studies; data collection and study appraisal; and synthesis and findings. From these four domains an overall summary of the risk of bias (ROB) is generated.
Results: From a random sample of 516 SRs, 87 (17%) prognostic reviews were identified. Most were SRs of one or more prognostic factors of disease (79, 91%). Two (2%) were classified as fundamental prognostic research, four (5%) considered the development, validation or impact of prognostic models. None considered the use of prognostic information to tailor treatment decisions and two (2%) covered multiple categories. Prognostic SRs were identified in thirteen disease areas with the majority relating to cancer (26, 30%) followed by mental health (10, 11%) and circulatory diseases (10, 11%). Twenty-two countries were represented in the sample with China contributing the most SRs (21, 24%) followed by the USA (14, 16%) and the UK (11, 13%). Generally, SRs were at high ROB across all four domains with domain 2 (identification and selection of studies) being the weakest (85% at high or unclear ROB). Overall, just seven SRs (8.5%) were at low ROB. The main areas of concern were not reporting a comprehensive search strategy, restrictions on sources of information and an inadequate assessment of the quality of the included studies.
Conclusions: SRs of prognostic studies are increasing in number but according to our random sample, despite extensive guidance available, the majority are at high ROB. The literature is dominated by cancer and most SRs investigated prognostic factors rather than models. We aim to use this research as a starting point for further exploration of trends in prognostic SRs.
Patient or healthcare consumer involvement: Whilst no healthcare consumers were involved, gaining a better understanding of the methodology of prognostic research should ultimately result in better healthcare outcomes for patients.