Pimonidazole labelling and response to fractionated irradiation of five human squamous cell carcinoma (hSCC) lines in nude mice: the need for a multivariate approach in biomarker studies.

OBJECTIVE: To investigate the influence on local control after fractionated radiotherapy of hypoxia measured in unirradiated tumours using the hypoxic marker Pimonidazole, using multivariate approaches. MATERIAL AND METHODS: Five human squamous cell carcinoma lines (FaDu, UT-SCC-15, UT-SCC-14, XF354, and UT-SCC-5) were transplanted subcutaneously into the right hind-leg of NMRI nude mice. Histological material was collected from 60 unirradiated tumours after injection of Pimonidazole. The relative hypoxic area within the viable tumour area (Pimonidazole hypoxic fraction, pHF) was determined in seven serial 10 microm cross-sections per tumour by fluorescence microscopy and computerized image analysis. Local tumour control was evaluated in a total of 399 irradiated tumours at 120 days after 30 fractions given within 6 weeks with total doses between 30 and 115 Gy. RESULTS: Tumour lines showed pronounced heterogeneity in both pHF and TCD50. Mean pHF values varied between 5% and 37%, TCD50 values between 47 and 130 Gy. A Cox Proportional Hazards model of time to recurrence with two covariates, dose and pHF, yielded significant contributions of both parameters on local control (p < 0.005) but violated the proportional hazards assumption, suggesting that other factors also influence tumour control. Introduction of histological grade as an example of a confounding factor into the model improved the fit significantly. Local control rates decreased with increasing pHF and this effect was more pronounced at higher doses. CONCLUSIONS: This study confirms that tumour hypoxia measured using Pimonidazole in untreated tumours is a significant determinant of local control after fractionated irradiation. The data support the use of multivariate approaches for the evaluation of a single prognostic biomarker such as Pimonidazole, and more generally, suggest that they are required to establish accurate prognostic factors for tumour response.