ESTRO 2024 - Abstract Book

S2411

Clinical - Urology

ESTRO 2024

Randomised controlled trials (RCTs) are the gold standard for testing the efficacy of new interventions. Radiotherapy is associated with both short term toxicities and late side effects and the latter can be dose limiting. The collection of late toxicity data in radiotherapy RCTs is therefore important in quantifying the longer term risks and benefits of new radiotherapy interventions. Typically, collection of late toxicity data in an RCT requires extensive input from both patients and clinicians to complete case report forms (CRFs) over many years. Linkage to healthcare datasets that are collected routinely across the United Kingdom has the potential to be a more time- and cost-effective method of obtaining long term follow up data in large RCTs whilst retaining reliability and accuracy of data [1]. The objective of this study was to examine whether routine healthcare data was a sufficient proxy for long term toxicity follow up within the CHHiP trial (ISRCTN97182923) using 10-year co-morbidity data collected as part of trial follow-up.

Material/Methods:

Data from the CHHiP 10-year co-morbidity CRF was compared to national (England) Hospital Episode Statistics (HES) datasets. As the 10 year co-morbidity form was designed to collect information on events occurring between years 6 and 10 of follow up, HES data was requested for this period.

ICD-10, OPCS-4.9 and A&E treatment or diagnosis codes from HES datasets that could be suggestive of the outcome of interest in the CRF question were reviewed.

For all eight CRF questions, patients were grouped into three cohorts: those with HES events alone, those with CHHiP CRF reported events alone and those who had matching events in HES and CRF data. Where a patient had the outcome of interest in both HES and CRF data the dates and any further information (ie fracture site) was reviewed where applicable to ascertain whether the same event was identified. Where there were enough events to enable analysis, the frequency of events (as determined by HES and/or by CRF) in each randomised treatment group was compared to ascertain whether using a particular dataset would have changed the interpretation of late-toxicity data from the trial.

Results:

CHHiP recruited 3216 patients; 2748 from England. 2653 (96%) of these could be linked to the HES dataset. Of the 2653 linked records, 1892 (71%) had a returned 10-year co-morbidity CRF form, and the analysis was therefore performed on those patients. 36 patients were recorded as having a fracture on the CHHiP trial CRF. Of those 23/36 also had a fracture recorded in HES datasets. There were 108 fractures recorded in HES data that were not recorded in CHHiP trial CRF (table 1). When compared between trial arms the use of HES data increased the prevalence of fractures but there remained no statistically significant difference between the trial arms. 167 patients were recorded as having a flexible sigmoidoscopy in the CHHiP trial CRF. Of these 53/167 were also recorded as having a flexible sigmoidoscopy in the HES datasets. 130 patients had a flexible sigmoidoscopy recorded in HES datasets alone. Using HES data increased the prevalence of flexible sigmoidoscopy but there remained no statistically significant difference between the trial arms.