Evaluate the diagnostic accuracy of median nerve cross-sectional area (CSA) to determine the severity of carpal tunnel syndrome (CTS) vs the presence of CTS across existing electrodiagnostic-based (EDX) classification systems. Retrospective analysis of cross-sectional patient data. Receiver operating characteristic (ROC) analysis was used to determine CSA cutoff values and associated diagnostic likelihood ratios for all consolidated and binary EDX-based classifications of CTS severity. Identification of CSA cutoff values associated with likelihood ratios capable of achieving conclusive (but at least moderate) shifts in diagnostic probability. Binary categorizations of CTS (ie, "Normal" vs "Absent") were statistically superior to consolidated categorizations of CTS severity (ie, "Normal," "Mild," "Moderate," or "Severe"). Binary categorizations established consistent CSA cutoff values across all EDX-based classifications examined and achieved conclusive shifts in diagnostic probability based on the following values of distal CSA or delta CSA: <7 or <1 mm2 to rule out and >13 or >7 mm2 to rule in CTS, respectively. Additionally, the following values of distal CSA and delta CSA may be used in certain circumstances because they produce only small shifts in diagnostic probability: ≤10 or ≤3 mm2 to rule out and ≥11 or ≥4 mm2 to rule in CTS, respectively. Using median nerve CSA to categorize the severity of CTS is not recommended based on lack of consistent and meaningful shifts in diagnostic probability. Rather, binary categorizations to rule out or rule in CTS based on the proposed CSA cutoff values consistently provided conclusive shifts in diagnostic probability across all EDX-based classifications examined.