Give me the ‘C’! How to define the comparator in a prognostic factor systematic review

Article type
Authors
Stallings E1, Jimenez D2, Muriel A3, Zamora J3, Fernandez-Felix B3, Quezada C2, Lopez-Alcalde J4
1Clinical Biostatistics Unit, Hospital Universitario Ramón y Cajal; Cochrane Madrid; CIBERESP
2Respiratory Department, Ramón y Cajal Hospital and Universidad de Alcalá IRYCIS, Madrid
3Clinical Biostatistics Unit, Hospital Universitario Ramón y Cajal (IRYCIS); CIBER Epidemiology and Public Health (CIBERESP)
4Cochrane Madrid / Universidad Francisco de Vitoria, Madrid; Instituto Ramón y Cajal de Investigación Sanitaria (IRYCIS), Madrid; Institute for Complementary and Integrative Medicine, University Hospital Zurich and University of Zurich; CIBERESP
Abstract
Background:
Our team is working on several prognostic factor systematic reviews (SRs). A SR aiming to determine the independent prognostic significance of a factor must come from a structured question (population, index factor, comparator, outcomes, timeframe and study design). The definition of the comparator requires identification of additional prognostic factors for which the prognostic association should be adjusted for. Ideally, these key additional prognostic factors should be defined at the protocol stage of the systematic review.
Objectives:
To describe the procedure that we followed to select the additional prognostic factors in the Cochrane review protocol titled “Sex as a prognostic factor in patients with acute symptomatic pulmonary embolism”.
Methods:
We carried out a bibliographic search in PubMed and Embase to identify prognostic factors in acute pulmonary embolism (PE). This search retrieved 6 factors, which we compiled in GRADEpro GDT. We sent the list to the systematic review team. In this first stage, the team commented on the factors already listed or added new ones. At this stage, a total of 24 factors were compiled.
In the second stage, the review team prioritised these factors, ranking them from 1 to 9 (with 1 being of least importance and 9 the highest). There was also the option to choose “Unknown”. Once all the team members had finished ranking, they were asked to confirm the list of factors and the order in which they had been prioritised.
Results:
The additional prognostic factors were classified into three groups: high priority (5 factors), low priority (9 factors), and excluded (10 factors). The high priority factors chosen were: immobilization history, history of surgery, history of recent bleeding, PESI score, and simplified PESI score.
Strengths: 1) our approach is transparent; 2) the process is straightforward in GRADEPro-GDT and doesn’t require presential meetings; 3) this process highlights the need to define an evidence-based procedure to define the list of additional confounders, which may be applied to any SR including non-randomised designs.
Limitations: 1) the criteria to define the relevance of the additional prognostic factors relied on clinical judgement only; however, there should be an evidence-based approach in place: for example, the additional adjustment factors should be statistically associated with both the outcome and the prognostic factor, and should not lie on a direct pathway between the prognostic factor and the outcome; 2) the maximum number of additional prognostic factors is not defined and to limit the number is not simple.
Conclusions:
We applied a transparent procedure for selecting additional prognostic factors to consider in a prognostic factor SR. This procedure can be applied to SRs of prognostic factor studies in the future. However, more research is needed to define the criteria on which to base the decisions to select the additional prognostic factors or the maximum number of factors to select.
Patient or healthcare consumer involvement:
None foreseen