Abstract
BACKGROUND: Invasive intra-arterial blood pressure (IABP) monitoring using an arterial catheter carries both risks and practical limitations. Noninvasive blood pressure (NIBP) monitoring using the widely adopted oscillometric automated brachial cuff is performed in nearly all patients in acute care settings at some point, yet its reliability remains in question. Conventional statistical approaches to assessing the global trueness (bias) and precision (standard deviation, SD) of NIBP compared to IABP may not fully capture the risk of harm posed by NIBP's measurement errors to patients, nor its ability to detect blood pressure (BP) values above or below critical thresholds. Moreover, risk factors for poor performance of NIBP warrant further investigation. We will perform a meta-analysis using individual participant data (IPD) to evaluate the clinical suitability of NIBP in acute care settings. METHODS: We will search the Cochrane Library, MEDLINE, and Scopus(™) databases to identify relevant peer-reviewed studies published in full-text English or French from 2000 to June 20, 2024. Studies will be included if they compare brachial cuff NIBP measurements with simultaneous radial, femoral, brachial, or pedal intra-arterial IABP measurements in adult patients in acute care settings. Retrospective studies will be excluded. Authors will be contacted to request IPD. For mean and systolic BP, we will assess the risk of harm associated with measurement error (via a dedicated error grid). As secondary objectives, we will evaluate (i) the bias between NIBP and IABP measurements (pooled estimate, SD, and limits of agreement), (ii) the precision (via within-patient SD, calculated for patients who had multiple paired measurements taken closely in time during stable periods), (iii) the ability of NIBP to detect IABP above or below critical thresholds and to identify response to therapy (via the area under the receiver operating characteristic curve), and (iv) the covariates associated with each of the abovementioned outcomes and with poor performance of NIBP. One-stage generalized linear mixed-effects models will be applied for the analyses. A meta-analysis combining IPD and aggregated data will also be conducted using a two-stage approach. Risk-of-bias assessment will follow the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) guidelines. For each outcome, a subgroup of particular interest will consist of studies with a low risk of bias. No funding is required. DISCUSSION: This meta-analysis will provide actionable insights to optimize BP monitoring strategies. TRIAL REGISTRATION: PROSPERO CRD42021233707.