A demonstration of an off-chip capacitance array sensor with a limit of detection of 1 μM trimethylamine N-oxide (TMAO) to diagnose a chronic metabolism disease in urine is presented. The improved Cole−Cole model is employed to determine the parameters of R _catalyzed , C _catalyzed , and R p_catalyzed , enabling the prediction of the catalytic resistance of enzyme, reduction effects of the analyte, and characterize the small signal alternating current properties of ionic strength caused by catalysis. Based on the standard solutions, we investigate the effects of pixel geometry parameters, driving electrode width, and sensing electrode width on the electrical field change of the off-chip capacitance sensor; the proposed off-chip sensor with readout system-on-chip exhibits a high sensitivity of 21 analog-to-digital converter counts/μM TMAO (or 2.5 mV/μM TMAO), response time of 1 s, repetition of 98.9%, and drift over time of 0.5 mV. The proposed off-chip sensor effectively discriminates TMAO in a phosphate-buffered saline solution based on minute changes in capacitance induced by the TorA enzyme, resulting in a discernible 2.15% distinction. These measurements have been successfully corroborated using the conventional cyclic voltammetry method, demonstrating a mere 0.024% variance. The off-chip sensor is crafted with a specific focus on detecting TMAO, achieved by excluding any reduction reactions between the TMAO-specific enzyme TorA and the compounds creatine and creatinine present in urine. This deliberate omission ensures that the sensor's attention remains solely on TMAO, thereby enhancing its precision in achieving accurate and reliable TMAO detection.