Article type
Abstract
Background: Clinical practice guidelines (CPG) should be based on the best scientific evidence, however for some interventions clinical randomised trials (RCT) are not available. Modelling studies is an alternative to evaluate benefits and harms in this scenario. In the context of a breast-cancer screening guideline, we conducted a systematic review of modelling and observational studies.
Objectives: To develop recommendations about the optimal breast cancer screening interval using the GRADE system.
Methods: We searched Pubmed and Embase and included modelling and observational studies which evaluated more than one mammography screening interval in average risk women. We developed evidence profiles to synthesise the evidence about effects and rated the certainty of evidence per outcome. We prioritise observational evidence when the certainty was at least similar to modeling evidence. For modelling evidence we used the ISPOR-AMCP-NPC questionnaire to assess credibility (risk of bias) and applied the rest of the GRADE domains (inconsistency, indirectness and imprecision).
Results: A total of 10 observational and nine modelling studies were included (Figure 1). Modelling studies predominantly used microsimulation technique, and were estimated for a United States population; results showed heterogeneity due to differences in assumptions and population data (Table 1 and 2). In general, annual screening offered the most number of breast cancer deaths averted but also higher harms. The tradeoff between benefits and harms was less positive in the 40 to 49 years strata than in the 50 to 69 years’ strata. Evidence was downgrade to very low quality due mainly to indirectness.
Conclusions: In the face of RCTs unavailability, the incorporation of modelling and observational evidence is a potential strategy that should be considered. Although some CPG in cancer screening have previously used this kind of evidence, the experience is still limited. A more explicit GRADE guidance is required for when to use and how to integrate modeling evidence into recommendations.
Objectives: To develop recommendations about the optimal breast cancer screening interval using the GRADE system.
Methods: We searched Pubmed and Embase and included modelling and observational studies which evaluated more than one mammography screening interval in average risk women. We developed evidence profiles to synthesise the evidence about effects and rated the certainty of evidence per outcome. We prioritise observational evidence when the certainty was at least similar to modeling evidence. For modelling evidence we used the ISPOR-AMCP-NPC questionnaire to assess credibility (risk of bias) and applied the rest of the GRADE domains (inconsistency, indirectness and imprecision).
Results: A total of 10 observational and nine modelling studies were included (Figure 1). Modelling studies predominantly used microsimulation technique, and were estimated for a United States population; results showed heterogeneity due to differences in assumptions and population data (Table 1 and 2). In general, annual screening offered the most number of breast cancer deaths averted but also higher harms. The tradeoff between benefits and harms was less positive in the 40 to 49 years strata than in the 50 to 69 years’ strata. Evidence was downgrade to very low quality due mainly to indirectness.
Conclusions: In the face of RCTs unavailability, the incorporation of modelling and observational evidence is a potential strategy that should be considered. Although some CPG in cancer screening have previously used this kind of evidence, the experience is still limited. A more explicit GRADE guidance is required for when to use and how to integrate modeling evidence into recommendations.