Table 3 Hypotheses on discrepancies between cohort studies and RCTs

Effect modification of the relationship between BP and health outcomes by frailty or other indicators of poor health

The predictive effect of BP on mortality and adverse health outcomes might be reversed by age-related frailty or other indicators of poor health (e.g., multimorbidity) [11]. Several cohort studies, which enrolled participants aged 60 years and over, have found that participants with no indicators of poor health with high BP had higher mortality rates and worse health outcomes, while participants with indicators of poor health with high BP had lower mortality rates and better health outcomes. The association may be modified by frailty [58, 59], limitations in cognitive and physical functioning [45, 48], and multimorbidity [49].

Confounding and reverse causality

Confounding: the relationship can be confounded by unmeasured factors, which have an effect on both BP and the risk of adverse health outcomes. Reverse causality: some conditions, which can be initially caused by high BP, evolve to become the cause of low BP [42].

Patient selection in clinical trials

RCTs might select participants in better health, with fewer comorbidities, and with a longer life expectancy than participants in population-based cohort studies, with the latter being more representative of the general population. For instance, HYVET and SPRINT trials have well-defined and restrictive eligibility criteria for participants, who are healthier than the general population of the same age [10]. Post-randomization confounding may also bias results of trials [62].