OBJECTIVES: Randomisation can be used as an instrumental variable (IV) to account for unmeasured confounding when seeking to assess the impact of non-compliance with treatment allocation in a randomised trial. We present and compare different methods to calculate the treatment effect on a binary outcome as a rate ratio in a randomised surgical trial.
STUDY DESIGN AND SETTING: The effectiveness of peeling versus not peeling the internal limiting membrane of the retina as part of the surgery for a full thickness macular hole. We compared IV based estimates (non-parametric causal bound, and two stage residual inclusion approach [2SRI] with standard treatment effect measures (intention to treat [ITT], per protocol [PP] and treatment received [TR]). Compliance was defined in two ways (initial and up to time point of interest). Poisson regression was used for the model based approaches with robust standard errors to calculate the risk ratio with 95% confidence intervals.
RESULTS: Results were similar for 1-month macular hole status across methods. For 3- and 6-month macular hole status, non-parametric causal bounds provided a narrower range of uncertainty than other methods, though still had substantial imprecision. For 3-month macular hole status, the TR estimate was substantially different from the other point estimates.
CONCLUSION: Non-parametric causal bound approaches are a useful addition to an IV estimation approach, which tend to have large levels of uncertainty. Methods which allow risk ratios to be calculated when addressing non-compliance in randomised trials exist and may be superior to standard estimates. Further research is needed to explore the properties of different IV methods in a broad range of RCT scenarios.
Bibliographical noteOpen Access funded by Medical Research Council
Acknowledgments: Jonathan Cook held an MRC personal fellowship (reference: G1002292) while some of this work was undertaken. The authors would like to thank the FILMS trial team (Noemi Lois, Jennifer Burr, John Norrie, Luke Vale, Jonathan Cook, Alison McDonald, Charles Boachie, Laura Ternent, and Gladys McPherson) and the FILMS study group (Hatem Atta, Stephen Beatty, Catherine Cleary, Andrew Dick, John Ellis, John Forrester, Carl Groenewald, Richard Haynes, Henrich Heimann, Muhammad Irfan Khan, Dara Kilmartin, Noemi Lois, John Murdoch, Asif Orakzai, CK Patel, Ian Pearce, Tarik Saddik, David Steel, David Wong, Charles Cottriall, Cherry Daly, Laura Duncan, Karon McEwing, Sarah Muir, Anita Murphy, Stan Keys, Lynda Lindsell, and Valerie Tompkin, Terri Ainley, Victor Beatty, Gillian Bennerson, Anne Bolton, Jon Brett, Alison Farrow, Ronnie Jackson, Tony Johnston, Marie Kinsella, Stephen Neilson, Hugh Nolan, Sarah Stanley, Jim Talbot, and Ayyakkawnu Manivannan) for the data set and Charles Boachie in particular for his previous analysis of the main data set. FILMS was funded by the Chief Scientist Office, Scotland (reference: CZH/4/235). The authors would like to thank the reviewers for helpful comments, which have improved the manuscript.
- Journal Article
- instrumental variable
- causal modelling
- risk ratio