A comparison of the effectiveness of different treatment regimens for pancreatic cancer using English cancer registry data

Large amounts of data are collected on cancer patients in the NHS, held by the National Cancer Registration and Analysis Service (NCRAS). Data are collected on patient and tumour characteristics, treatments, and can be linked to hospital episode statistics. Usually new cancer treatments are investigated in randomised controlled trials (RCTs), which are widely considered to represent the gold standard approach for comparing interventions. However, sometimes it is not possible to run an RCT, due to feasibility or ethical issues. In addition, RCTs often have strict and restrictive eligibility criteria. Whilst RCTs might tell us about the comparative effectiveness of treatments in highly selected trial populations, they are less useful for investigating comparative effectiveness in more general populations. Therefore, it is important to investigate the use of NCRAS data as a resource for estimating the comparative effectiveness of cancer treatments in the “real world”, that is, under routine conditions. This project aims to investigate whether or not English cancer registry data is sufficient for deriving valid causal estimates of the comparative effectiveness of different cancer treatments given in the NHS.

This document provides a protocol for carrying out four Target Trial Emulations (TTE), using NCRAS data. Each TTE seeks to emulate as closely as possible an existing RCT investigating treatments for pancreatic cancer. We describe each TTE in detail, and specify agreement criteria that will be used to evaluate the success of each emulation. This study will provide valuable evidence on whether it is possible to derive robust and valid causal estimates of comparative effectiveness of cancer treatments given in the NHS. If we are able to successfully emulate existing RCTs, our study will provide evidence that obtaining such estimates is possible, and will provide the basis for designing analyses that seek to answer questions not addressed by RCTs. If we are not able to successfully emulate existing RCTs, our study will seek to identify key weaknesses in the registry datasets, with the intention of determining how these datasets could be improved. Therefore, our study has the potential to provide valuable insights for healthcare decision-makers, clinicians, and patients.