Glucose sensors based on immobilized glucose oxidase and hydrogen peroxide detection at a platinum base electrode were constructed and studied before, during and after implantation into the subcutaneous tissue of 11 non-diabetic subjects. A 75-g oral glucose load was given to elevate the blood glucose concentration. Seven of 14 sensors responded to the oral glucose administration with an increase in current and the output of the remainder was unchanged by the glucose load. Apparent subcutaneous glucose levels calculated from the pre-implantation calibration were a mean 58% of the plasma glucose values at baseline. A two-point in vivo calibration using paired current and glucose readings at baseline and at the maximum glucose and current after glucose ingestion showed a significantly reduced sensitivity in vivo compared with pre-implantation values (mean +/- SEM 52 +/- 21.5 vs 369 +/- 127 pA/mmol-1 per litre, P = 0.003). Recalibration of the subcutaneous glucose concentrations using the in vivo calibration sensitivity and extrapolated background current (I0) gave values similar to those in plasma. The sensitivity of five sensors recalibrated in vitro after explantation was also reduced compared with pre-implantation levels and not significantly different from the in vivo characteristics. Responding and non-responding sensors did not differ with respect to preimplantation I0, sensitivity or response time. However, provisional examination of some explanted sensors by scanning electron microscopy showed coating by cellular and other amorphous material in the non-functioning electrodes. We conclude that the sensitivity of glucose sensors of this design is markedly reduced, sometimes to zero, on implantation in the subcutaneous tissue of humans.(ABSTRACT TRUNCATED AT 250 WORDS)