Personalized prophylactic anticoagulation decision analysis in patients with membranous nephropathy

Abstract

Primary membranous nephropathy is associated with increased risk of venous thromboembolic events, which are inversely correlated with serum albumin levels. To evaluate the potential benefit of prophylactic anticoagulation (venous thromboembolic events prevented) relative to the risk (major bleeds), we constructed a Markov decision model. The venous thromboembolic event risk according to serum albumin was obtained from an inception cohort of 898 patients with primary membranous nephropathy. Risk estimates of hemorrhage were obtained from a systematic literature review. Benefit-to-risk ratios were predicted according to bleeding risk and serum albumin. This ratio increased with worsening hypoalbuminemia from 4.5:1 for an albumin under 3 g/dl to 13.1:1 for an albumin under 2 g/dl in patients at low bleeding risk. Patients at intermediate bleeding risk with an albumin under 2 g/dl have a moderately favorable benefit-to-risk ratio (under 5:1). Patients at high bleeding risk are unlikely to benefit from prophylactic anticoagulation regardless of albuminemia. Probabilistic sensitivity analysis, to account for uncertainty in risk estimates, confirmed these trends. From these data, we constructed a tool to estimate the likelihood of benefit based on an individual's bleeding risk profile, serum albumin level, and acceptable benefit-to-risk ratio (www.gntools.com). This tool provides an approach to the decision of prophylactic anticoagulation personalized to the individual's needs and adaptable to dynamic changes in health status and risk profile.

Figure 1. Hybrid Markov decision tree model

The square indicates a decision node in which the clinician may choose from either of two alternatives—prophylactic anticoagulation or observation and treatment of clinically apparent venous thromboembolic events (VTEs). The circles represent chance nodes in which events occur at random (i.e., outside the clinician’s control). The letter M in a circle represents a Markov node and the triangles represent terminal nodes or clinical end outcomes. AC, anticoagulation; ESRD, end-stage renal disease; NS, nephrotic syndrome; obs, observation; VTE_obs, VTE occurring in the observation strategy; VTE_{on AC}, VTE occurring while on prophylactic anticoagulation; VTE_{off AC}, VTE occurring subsequent to a major bleeding event.

Figure 2. Consideration of uncertainty: probabilistic sensitivity analysis

Monte Carlo simulation plots of benefit and risk at various levels of hypoalbuminemia; plots show the simulation results according to combinations of risk categories of major bleeding in the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) study and different serum albumin levels: (a) <3.0 g/dl, (b) <2.5 g/dl, and (c) <2.0 g/dl. Each graph includes three density plots representing high risk bleeding risk (upper, open circles), intermediate bleeding risk (middle, open diamonds), and low bleeding risk (lower, open triangles), respectively. Three benefit-to-risk threshold lines are shown for reference (2:1, 5:1, and 10:1). Simulation points lying to the right and below the threshold lines indicate that the benefit-to-risk ratio is larger than the threshold. For example, points lying below and to the right of the ‘10:1 ratio’ line have benefit-to-risk ratios >10:1.

Figure 3. Benefit–risk acceptability curves

Benefit–risk acceptability curves represent the probability (vertical axis) of prophylactic anticoagulation being beneficial at various levels of the benefit-to-risk ratio (horizontal axis). Major bleeding risk categories (low and intermediate) were derived from the ATRIA study. Each curve corresponds to each serum albumin level (from top to bottom,<2.0 to <3.0 g/dl, respectively). (a) Low risk of major bleeding, (b) intermediate risk of major bleeding, and (c) high risk of major bleeding.

Figure 4

Decision approach for prophylactic anticoagulation.