Influenza vaccine effectiveness in patients with chronic obstructive pulmonary disease: a cohort study emulating a randomised trial

Objectives Many observational studies of influenza vaccine effectiveness have faced issues of residual confounding and immortal time bias. We aimed to assess the effectiveness in people with chronic obstructive pulmonary disease (COPD), using a novel study design to address these methodological challenges. Methods We conducted a cohort study emulating a randomised trial within the primary care practices contributing data to England’s Clinical Practice Research Datalink. We identified the base cohort of patients diagnosed with COPD, aged 40 years and above, initiating an inhaled treatment. Those vaccinated between September 2014 and March 2020 were matched 1:1 with unvaccinated people in the week of vaccination and by time-conditional propensity score. The primary outcome was hospitalised influenza during epidemic periods. Secondary outcomes included a composite of acute exacerbations of COPD, pneumonia, and hospitalised influenza. Urinary tract infections during epidemic periods and the composite outcome outside of influenza seasons (May to August) were used to detect residual confounding. Results The emulated trial included 56 652 matched pairs of vaccinated and unvaccinated patients with COPD. The hazard ratio (HR) for hospitalised influenza was 0.61 (95% CI, 0.48–0.77) during epidemic periods. The HR for the composite secondary outcome was 0.89 (95% CI, 0.87–0.92) during epidemic periods and 0.99 (95% CI, 0.94–1.04) outside influenza seasons. The HR for urinary tract infection was 0.83 (95% CI, 0.77–0.90). Results were consistent in several sensitivity analyses, such as when excluding patients consulting for reasons likely precluding vaccination. Conclusions Influenza vaccine was associated with a reduced risk of hospitalised influenza in patients with COPD, but a reduced risk of urinary tract infection suggests important residual confounding. Although this study design addresses immortal time bias, methodological improvements are needed to better address residual confounding in cohort studies of the influenza vaccine. In the meantime, negative control outcomes should be assessed during influenza epidemics to better detect residual confounding.