ABSTRACT Background Little is known about the influence of sociodemographic factors on patient access to early phase cancer trials. The toxicity and efficacy of cancer drugs can vary according to sociodemographic factors, and these differences should be considered to ensure generalisability of results and equality of access. Method We conducted a review of patients referred to the early phase trials unit at our centre in the five years to 2012. Electronic records were studied for demographic and cancer-specific data. Socio-economic status was defined by the Index of Multiple Deprivation (IMD; 1 - least deprived, 5 - most deprived) recorded for the regional Cancer Registry population according to postal code. Multivariate analysis (adjusting for gender, age and tumour type) was performed comparing 10,784 incident cancer cases in south east London with the patients referred to our unit, and with those enrolled in a trial. Results 430 patients (195 female) were referred for consideration of an early phase trial, with a median age of 62 years (range: 22-86). Ethnicity was 74% white. Univariate analysis of ethnicity suggested the non-white population was less likely to be recruited (OR 0.48, 95% CI 0.26-0.88), but this relationship was lost with adjustment for age, gender, cancer type and IMD. Multivariate analysis showed that referral was less likely for patients in the more deprived quintiles (IMD-5: OR 0.53, 95% CI 0.38-0.74). However, once referred to the unit, enrollment in a trial was not affected by IMD (IMD-5: OR 0.81, 95% CI 0.40-1.63). Conclusion We show for the first time that social deprivation affects referral to an early phase cancer trials unit. The least deprived patients are almost twice as likely to be referred to the trial unit compared with the most deprived. This may be because patients in the higher deprivation quintiles are less suitable for a trial, for example due to comorbidities, or because of inequalities that could be addressed with patient or referrer education. Disclosure All authors have declared no conflicts of interest.