We sought to corroborate geographical differences in hepatitis C virus (HCV) prevalence and assess whether these can be explained by differences in injecting risk behaviour. A community recruited interview survey of 1058 injecting drug users (IDU) - including a blood spot specimen for antibody testing - was undertaken in seven cities in England. HCV prevalence varied from 27% to 74% across sites (chi(2)(6) = 115.3, P < 0.001). There was a significant variation in crack-injection, prison history, injecting frequency, homelessness, groin injecting, syringe reuse and sharing between the sites. Adjustment for clustering by site and other covariates attenuated the odds ratios (OR) for most variables: e.g. crack injection changed from an unadjusted OR of >2 to an adjusted OR of 1.4 (95% CI 0.9-2.0). Remaining significant covariates included: homelessness (OR 2.2; 1.4-3.6); ever imprisonment (OR 1.7; 1.2-2.5); syringe sharing >18 months ago (OR 2.0; 1.3-3.0); injecting duration and age. Introducing site as a second level variable did not reach significance (P = 0.10). HCV prevalence among IDU reporting 'never sharing' was 48%. Geographical variation in HCV prevalence remains poorly explained, but should be the key focus of our surveillance effort. Measures of sharing and their interpretation require greater scrutiny.