Results of a retrospective linkage of TACT (CRUK 01/001) breast cancer trial data and the National Cancer Data Repository


Session type:

L. Kilburn1, M. Aresu1, J. Banerji1, J. Bliss1
1Institute of Cancer Research, London, UK


Randomised clinical trials (RCTs) are the gold standard method to determine the value of cancer treatments. They are, however, expensive to run, particularly for long-term follow-up of participants. The National Cancer Data Repository (NCDR) combines hospital episode statistics (HES) with registry data for English patients. Tracking trial participants via routine data such as NCDR may be more cost-effective for follow-up than traditional methods. This project assesses the potential of the NCDR to inform breast cancer clinical trials using mature data available from the TACT early breast cancer trial.


Data collected via the TACT trial were matched initially to the NCDR using the patients' NHS numbers. Further linkage was conducted using date of birth (dob) and patient name/initials where available. Demographics, clinical characteristics and outcome data were assessed for agreement and completeness. Overall survival was compared between NCDR and TACT.


3,347 TACT patients' data were sent to NCDR to undergo initial matching. 3,058 (91.4%) were successfully matched using NHS number alone and 3,047 (91.0%) matched using NHS number, dob and name/initials. Extensive cleaning was required for some registry data fields, for example cause of death, whilst others had large amounts of missing data, e.g. tumour sizes (25%). Other data had high levels of matching such as dob (96%), date of death (88%) and laterality (97%). There was no evidence of differential survival rates as calculated from NCDR and TACT (5-year survival: TACT=82.8% 95%CI (81.4, 84.1); NCDR=83.0% 95%CI (81.7, 84.3)).


Whilst for some data fields, agreement was high, data quality and completeness would require improvement especially in registry based data before such data could be used for RCTs. Definition of new datasets, including COSD, is welcomed. Prospective validation of such datasets is required before RCTs can confidently switch patient follow-up to utilise routinely collected NHS based data.