A suspicion index tool to aid the diagnosis and treatment of ASMD

The diagnosis of acid sphingomyelinase deficiency (ASMD, Niemann Pick Type A, A/B, B) is frequently delayed by years, because of its heterogeneous and often unspecific clinical features. The involvement of multiple organs including the musculoskeletal and central nervous systems, poses a challenge for accurate diagnosis. To address this, we developed a suspicion index tool (SIT) for healthcare professionals to enable early and accurate diagnosis of ASMD. Our methodological approach encompasses five key steps: (i) literature research on ASMD symptomatology, (ii) retrospective expert chart review of international ASMD centers, (iii) retrospective statistical analysis, (iv) development of an individual risk prediction score via random forest regression, and multinomial modeling, (v) internal validation of the tool via bootstrap resampling. Data were collected from 908 patients (48 cases, 52 controls, and 808 non-cases) from eight expert centers. Visceral symptoms emerged as strong indicators of ASMD, particularly isolated unexplained splenomegaly (100% of cases vs. 71% of controls and 0.4% of non-cases) and hepatomegaly (92% of cases vs. 56% of controls and 0.4% of non-cases). Respiratory symptoms, thrombocytopenia, and hypercholesterolemia were also identified as significant indicators. These variables were selected for inclusion in the final SIT using a best subset selection algorithm. Each variable composition was evaluated via extensive repetitions. Additionally, expert input was sought to assess the significance of selected variables. The SIT demonstrated superior accuracy, sensitivity, specificity, and internal validity, confirming its reliability. The SIT is currently under development as a web-based platform for facilitating the diagnosis of ASMD and other treatable diseases in at-risk populations. Early diagnosis of ASMD is crucial but often delayed. Our newly developed innovative diagnostic tool, which leverages a large patient dataset and advanced statistical and machine learning techniques, can improve the accuracy and reliability of ASMD diagnosis by considering multiple symptoms and enabling patient participation.