Survival estimates in European cystic fibrosis patients and the impact of socioeconomic factors: a retrospective registry cohort study

Discussion

We have shown that median survival in patients with CF across Europe is comparable to that of Canada and the USA, and that there is variation across Europe that is associated with socioeconomic factors.

Survival for patients with CF is variable and is influenced by factors including background CFTR genetics and environmental exposures [2]. CFTR genotypes with at least one class IV–V CFTR mutation have a milder phenotype and better survival [15, 16]. Likewise, environmental factors such as acquisition of Pseudomonas aeruginosa [17], Staphylococcus aureus and Burkholderia cepacia complex [18] also influence mortality. In the USA, there is a clear association between individual-level socioeconomic status and CF outcomes, with absence of private medical insurance and lower median income independently associated with higher mortality [19, 20]. This relationship between socioeconomic status and survival in CF is multifactorial, with access to healthcare, education, adherence and expectations all contributing to differences in outcomes [21]. In Europe, McCormick et al. [6], using the European CF Demographics Registry dataset (a precursor of ECFSPR), demonstrated differences in demographics across Europe with a median patient age of 17.0 years in European Union (EU) countries compared with a median patient age of 12.1 years in non-EU countries. The proportion of patients aged >40 years was twice as high in EU countries than non-EU countries, raising concerns about underdiagnosis of CF and increased childhood mortality as a result of unequal access to specialist CF care and CF medicines. This was consistent with the earlier work of Fogarty et al. [5] who also found differences in median age of death for CF patients across countries, which they attributed to possible underdiagnosis and diagnostic misclassification of CF as well as socioeconomic factors.

One of the challenges of comparing differences in survival across countries has been differences in statistical methodology in single-country registry annual reports [22, 23]. In a recent study looking at survival in the US and Canadian CF patient registries, using the same methodology for survival analysis [24], there was an almost 10-year difference in median survival that has been increasing since 2005 [3]. Socioeconomic factors, nutrition and access to lung transplantation were all considered to influence this difference in survival [4]. Median survival in CF patients was 40.6 years in the USA compared with 50.9 years in Canada. The median survival estimated in our ECFSPR study, using a similar statistical methodology, was 51.7 years. However, a limitation of our study is that many European patients in the ECFSPR have limited data after lung transplant as many transplant centres are not enrolled in the ECFSPR. This results in individuals tending to be lost to follow-up at the time of transplant, which is likely to increase the survival estimates. In the USA–Canada study, censoring at time of transplant resulted in increased median survival in the USA to 44.0 years and to 57.1 years in Canada [3]. Our median survival censoring at transplant of 56.8 years lies between these estimates for the USA and Canada, which is likely to be a more accurate comparison.

The difference between median survival including post-transplantation follow-up (51.7 years) and using the composite outcome of death or transplant (38.5 years) highlights the impact of transplantation and the improved survival after transplantation [25]. This difference may be due to uncounted deaths in patients lost to follow-up post-transplant, as well as differences in access to transplantation in some countries reflected by a highly different percentage of transplanted patients among those seen each year, which varied between >12% in the UK and 0% in some Eastern European countries [9]. There were also differences in median survival when we limited the cohort to those homozygous for F508del, which is similar to other reports [15]. The distribution of F508del differs across European countries [9] and because of this, the influence of socioeconomic factors on survival was adjusted for CFTR genotype to account for differences in genotype frequencies in countries with lower socioeconomic measures.

Our study also demonstrates that survival outcomes vary depending on different socioeconomic factors. Studies of socioeconomic status and CF outcomes in the USA have shown that medical insurance status [20] and median household income [19] are both independently associated with difference in CF survival outcomes, even within a country with a high GNI. In the UK, a validated deprivation score was associated with poorer outcomes, including increased infection with P. aeruginosa and decreased access to and use of CF medications, all of which are associated with reduced CF survival [26]. This is the first study in Europe to quantify the association between national socioeconomic factors and survival, and shows that countries with the lowest measures of healthcare spending have hazard rates for death that are almost twice that of countries with higher measures of healthcare spending. This increase in hazard with lower socioeconomic spending was consistent across three separate country-level socioeconomic metrics. Despite common European Standards of Care for CF and a national health insurance system in almost all European countries, access to care and medication varies widely across Europe, especially in Eastern Europe [27]. The association between socioeconomic factors and CF survival is not unexpected as standardised mortality from all causes differs across Europe (as shown in table 3), although the magnitude of effect in CF is greater than that seen for the general population and demonstrates the need for further research in this area within Europe.

There are a number of limitations to our study. Missing data and data quality are always challenging in studies using registry data. The analysis was limited to countries with a national registry and coverage of >80% of their CF population. It was assumed that missing data on covariates within countries were missing completely at random. By restricting our cohort to countries with a national registry, we assumed that the combined population is representative of that in Europe as a whole. The overall median survival age could be subject to bias if this is not the case. However, the findings relating to association between socioeconomic factors and survival would only be biased if the association between socioeconomic factors and survival differed in countries that were not a part of this study. All of the included national registries have rigorous approaches to data quality. This, in addition to the data quality requirements of the ECFSPR [28], increases the likelihood that the results are reliable. At the time of study completion, data from two large European countries (Germany and Spain) were not available and it is possible that the survival estimates may change with the inclusion of these large countries. This will require a further follow-up study. Likewise, it is important to note that these survival estimates for Europe cannot be extrapolated to individual patients with CF or to all European countries as the median estimates are influenced by survivorship in the larger European countries. The three largest countries (UK, France and Italy) contributed 23 849 patients (75%) and 1093 deaths (76%), indicating that the median survival largely reflects the median survival of these three countries. We chose not to weight the survival estimates by country population as we were studying regional differences and used a similar methodology to that used for Canada and the USA. Also, to ensure as accurate a survival estimate as possible, we restricted the cohort to include countries with the highest coverage and the most complete data. Future analysis including more countries, especially Eastern European countries, will be planned once the ECFSPR has sufficient data to do so. It is also worth noting that these estimates reflect a cohort of CF patients followed before the widespread availability of CFTR modulator therapy and future survival estimates may change as these highly effective novel CF therapies become more widely used across Europe.

Finally, a number of deaths may have been missing. It is anticipated that this number is low as most CF patients were attending CF centres who would generally know each patient's vital status, although we acknowledge that outcomes post-transplant may be incomplete. The absence of follow-up post-transplant in some countries limits the interpretation of the overall survival estimates. The median survival estimate may be biased due to the exclusion of post-transplant deaths. As seen in table 1, the proportion of deaths when censored at transplant compared with total deaths is highly variable across European countries. This is likely to be due to differences in access to transplant, post-transplant loss to follow-up in the registry and transplant centre survival rates in the different European countries. As all of these factors may influence the median survival estimate, attempts are underway to audit data quality and number of deaths as well as include post-transplant centre data in the ECFSPR. We also included transplant status as a time-dependent covariate in multivariable Cox regression analyses. However, this may be a mediator of the association between socioeconomic factors and mortality (e.g. if access to transplant is affected by socioeconomic factors). We also did not allow the hazard ratio for transplant to depend on time since transplant. Another potential source of bias is noninformative censoring. The models used assume censoring (due to loss to follow-up) is uninformative for the event of interest. Loss to follow-up rates were generally low, but it is possible that covariates not included in our model may have influenced differences in each country's loss to follow-up. Unfortunately, there is no way to formally test this. In the analysis in which we censor patients at transplant, the focus is on cause-specific hazards for pre-transplant mortality and so censoring at transplant is not considered a form of informative censoring.

In conclusion, this study demonstrates that median survival for patients with CF in Europe is comparable to that reported in the USA and Canada, and that survival across Europe is highly influenced by socioeconomic factors. A more detailed understanding of how these differences in socioeconomic factors lead to poorer survival is critical to improving outcomes for CF patients from European countries with lower healthcare spending.