Positron emission tomography (PET) provides non-invasive, in vivo, functionalimaging data. Using PET and [11Cmethyl]temozolomide radiolabelled in the 3Nmethyl position, these studies demonstrate pharmacokinetics safely in vivo in man at tracer doses (SOmcg). The biodistribution of [11Cmethyl]temozolomide in the normal extracranial organs is shown and penetration of the blood-brain barrier is confirmed with accumulation of tracer in primary brain tumours. Using area under the curve (AUC) calculations, tumour [llCmethyl]temozolomide exposure is significantly greater than that of normal brain (p < 0.001), although less tumour exposure is seen in the full-dose studies, (p < 0.002). Tumour perfusion does not affect tumour [11Cmethyl]temozolomide exposure, yet there is greater tumour extraction (p < 0.02) and retention (p < 0.05) of [11Cmethyl]temozolomide at 60 minutes. A relationship is seen between [11Cmethyl]temozolomide retention at 90 minutes in tumours with response duration and survival (p < 0.05). Using fluorine-18 radiolabelled fluorodeoxyglucose ([18F]FDG) which provides a measure of tissue glucose utilisation, pharmacodynamic endpoints can be related to tumour response. A reduction of > 25% in glucose utilisation of the tumour high uptake focus (MRGluHuF), within 2 weeks of starting temozolomide, is shown to correlate with patient clinical response. A < 25% change in MRGlumJF correlates with stable disease. The 3-compartment modelling (MRGlu) is better than standardised uptake values (SUV) at separating patients into responding and non-responding groups. Assessment of less invasive methods of generating the arterial input function necessary for 3-compartment modelling determined that although left ventricular sampling (n=l0) is a good surrogate method (r2=0.7940) a population-generated, arterial input function is better (r2=0.8789). In addition, it was demonstrated that it was 5 preferable to scale the population-generated input function using the arteriovenous [18F] concentration. Variability of < 4.5% (tumour) and < 6.1% (brain) is seen using arteriovenous scaling and 3-compartment modelling.