Vector autoregression: Useful in rare diseases?—Predicting organ response patterns in a rare case of secondary AA amyloidosis

Abstract

Background Statistical analyses of clinical data are a cornerstone in understanding pathomechanisms of disorders. In rare disorders, cross-sectional datasets of sufficient size are usually not available. Taking AA amyloidosis as an example of a life-threatening rare disorder resulting from of uncontrolled chronic inflammation, we propose techniques from time series analysis to predict organ response to treatment. The advantage of time-series analysis is that it solely relies on temporal variation and therefore allows analyzing organ response to treatment even when the cross-sectional dimension is small. Methods The joint temporal interdependence of inflammatory activity and organ response was modelled multivariately using vector autoregression (VAR) based on a unique 4.5 year spanning data set of routine laboratory, imaging data (e.g., 18F-Florbetaben-PET/CT) and functional investigations of a 68-year-old patient with multi-organ involvement of AA amyloidosis due to ongoing inflammatory activity of a malignant paraganglioma in stable disease for >20 years and excellent response to tocilizumab). Results VAR analysis showed that alterations in inflammatory activity forecasted alkaline phosphatase (AP). AP levels, but not inflammatory activity at the previous measurement time point predicted proteinuria. Conclusion We demonstrate the feasibility and value of time series analysis for obtaining clinically reliable information when the rarity of a disease prevents conventional prognostic modelling approaches. We illustrate the comparative utility of blood, functional and imaging markers to monitor the development and regression of AA amyloidosis.

Increasing contribution of the atmospheric vertical motion to precipitation in a warming climate

“…The VAR extends this univariate autoregression to the system of regressions where all variables are predicted by their own lagged values, all other variables, and their lagged values. This model is pioneered by Sims [35] and has been applied in estimating the impact of economic shocks [36,37], modeling brain network [38,39], and predicting disease [40,41].…”