Cystic fibrosis (CF) is a disease affecting over 10,000 people in the UK. It has no cure, but there are many treatments to help improve health. Randomised controlled trials are the gold standard for establishing treatment efficacy, but most trials for CF treatments have no more than one year of follow-up. In practice treatments are commonly used for many years, and it is therefore important to evaluate their long-term effectiveness. The UK CF Registry collects annual data on almost all people with CF in the UK. The overall aim of this work is to investigate how data from such registries can be harnessed to provide insights into the effects of long-term treatment use. My research illustrates the potential of registry data by investigating two CF treatments: DNase and ivacaftor. DNase is a common CF treatment and generally, once started, it continues to be used indefinitely. Despite this, no studies have investigated its long-term effects. Estimating these effects using registry data is difficult due to time-dependent confounding. I investigate five methods that can account for this: sequential conditional mean models, inverse probability weighting of marginal structural models (MSM), history-adjusted MSM, gcomputation formula and g-estimation of structural nested models. The performance of these methods is assessed through simulation studies, where it is shown that all methods perform similarly under correct model specification, suggesting that more than one method could be applied to assess consistency of results. My analysis of the UK CF Registry data suggests that DNase provides a step-change improvement in lung function only in individuals with ppFEV1 < 70% (e.g. for a person starting DNase with ppFEV1 of 20%, the one-year treatment effect was a 1.6% absolute difference in ppFEV1, 95% CI 0.4, 2.8). However, the slope of lung function decline over five years remained unchanged. Ivacaftor was introduced in the UK in 2012, but it is only available to people with a gating mutation. In this subgroup, it appears to be so beneficial that almost all eligible people are now receiving it. In this situation, it is difficult to estimate the treatment effect, because there are no eligible people not receiving treatment. Two possible comparator groups were identified: 1) those currently receiving ivacaftor, but using their data from years prior to its introduction, 2) those ineligible to receive ivacaftor due to their genotype. This work shows how analyses using negative controls can be used to assess the comparability of the different groups, and how differences between groups not due to treatment can be mitigated. Our analysis suggests that these two groups are comparable to people who are currently receiving ivacaftor, and the results of the analysis show that ivacaftor not only provides an initial step-change improvement in lung function (5.9% absolute difference in ppFEV1, 95% CI 4.7, 7.1), but also decrease the rate of lung function decline (0.5% absolute decrease in ppFEV1 decline per year, 95% CI 0.02, 1.0).