The outcome of cystic fibrosis (CF) patients has improved in recent decades, with mortality less than 5 percent during the first 10 years of life in patients treated with current multidisciplinary care 1 . However, disease severity remains variable among children, adolescents and adults 2 3 . In CF, severity is principally assessed by the decline in lung function because lung disease still remains the most common cause of death. Lung function in CF is almost universally calculated as percent predicted FEV₁ referenced against a healthy population 4 5 . This particular choice reflects a widely held opinion that achieving normality remains the ultimate goal of CF care. Irrespective of whether or not this aspiration is achievable, a limitation of the current ‘reference against the normal range’ approach is that it does not provide a ranking of an individual patient’s status relative to age and sex-matched CF peers. Similar considerations apply to nutrition, which must be adequate to maintain lung function 6 , or surrogate markers of disease severity calculated from semi-quantitative screening scales such as the Chrispin Norman Score that measures lung damage from chest radiography 7 .

There have been attempts to overcome such limitations. Kulich and coworkers converted the absolute FEV₁ into percentiles calculated from a registry of lung function values in North American CF patients 8 . In a related manner, McCormick and colleagues converted a chest x-ray severity score in childhood into population based percentiles 9 . Such self-referencing approaches, despite their calculation from cross sectional data, provide a reference base that allows the longitudinal tracking of CF disease outcome and informs on the relative position of a given patient against his or her peers.

As of now, the CF FEV₁ specific percentiles have of necessity been obtained from US CF patients. It remains to be established whether these data are appropriate for European patients with CF. Indeed, there are many differences not only in how CF care is organized (for example, coverage of neonatal screening, timely referral to reference centers, variation in standards of care) but also in the environment between EU nations and across the Atlantic. Furthermore Kulich analyzed data from 15 years ago (1994 to 2001) that would not account for recent progress in CF care. To mitigate against such issues, we combined data from the European Cystic Fibrosis Society Patient Registry (ECFSPR) and data from the current French CF Modifier Gene Study, to obtain CF reference percentile equations for lung function and body mass index that would apply to current European CF patients and would additionally provide the ranking of an individual patient’s status relative to age and sex-matched CF peers. As a second objective, we examined whether FEV₁ and BMI in CF patients differed between European countries and across two continents, Europe and North America (USA).

Achieving normality in lung function or nutritional status is the ultimate goal of CF care, even if the possibility of achieving this ideal remains a moot point given the multisystem nature of CF and the complexities of the variable genotype-phenotype relationship 17 . Using a healthy population as a reference (as is universal with FEV₁ pp) determines how far CF patients lie from normality. We propose here that additional information may be gleaned by referencing CF patients against age and sex matched CF peers. A large database of European cystic fibrosis patients was used to compute these reference percentiles and the software code is available from the authors on request.

As of today, such CF-specific reference- ranges have mainly been used for research purposes, for example to provide a quantitative phenotype for comparing patients 18 . However, they could also be useful in a given clinic to either directly compare the performance between CF centers or to help homogenize patients for inclusion in clinical trials. This disease-referencing approach could also be useful to physicians and patients to help visualize the status of a single patient relative to his or her peers. For example, a longitudinal plot could highlight unusual worsening of clinical conditions within European CF framework and help the physician on the need for more aggressive therapies.

One may worry that referencing CF patients relative to their peers could lessen physicians’ efforts to improve CF patients’ health status, if they were satisfied with their patients’ progress referenced against their peers. In other words, care must be taken in interpretation such that the CF specific percentiles should not be interpreted as “normative” values, but, as discussed above, as a reference providing additional information. Reassuringly, there was no indication that the availability of such equations from US patients since 2005 negatively impacted CF care as also evidenced by the approach of McCormick and colleagues who calculated similar centiles for chest X-ray scores. Furthermore, our local patient representatives, as required by our consents and ethical practices on the use of such data, when shown the approach, spontaneously remarked that they would like to know how they were performing relative to others in Europe. Future studies will have to determine the impact of providing both types of reference information, for example in the case of specific CFTR variants such as G551D-CFTR which now has a new therapy 19 .

Of necessity, it will be imperative to regularly update the CF specific reference equations to reflect CF care improvement. For example, it was reassuring that the European CF-specific FEV₁ percentiles were greater than those computed a few years ago from US CF patients 8 . This difference was more pronounced in the lower percentiles at all ages. It is unlikely that this difference results of differential mortality, as mortality curves are very similar for CF patients in large European countries and the US 20 21 . Better organization of CF care over time may have led to a larger number of less severe patients included in registries, therefore leading to improved overall performance. However, it remains equally likely that changes in the efficacy of CF care over time is an explanation for the observed differences given that almost one decade passed between data collection in the US and the European studies. Significant improvement in the survival and clinical status of CF patients has been achieved during this time, by earlier CF diagnosis, better nutritional support and mucus drainage, and better diagnosis and treatment of CF-related complications 22 . An argument in favor of the increased care efficacy is that the improvement was more pronounced in the patients with the poorest lung function. In addition, the median CF specific FEV₁ percentile calculated in today’s North-American patients using the Kulich equations tends to be above the expected 50^th percentile 18 .

Using the ECFSPR data permitted analysis of a large number of measurements obtained from several European countries. Participation to the ECFSPR is on a voluntary basis, and the coverage (i.e. the proportion of CF living population actually included in the registry) ranges from 15% to >99%, with 9 countries having coverage greater than 50%. Little selection bias is expected in countries with large coverage, while it may be significant in countries where participation is limited to some voluntary CF centers. In the latter case, the extent to which the reported patients’ characteristics are biased relative to the whole country CF population is unpredictable. However, in our analysis, the estimates were not substantially affected when we excluded countries with small coverage.

We computed the CF specific equations so that unselected data could be referenced against these curves. The FEV₁ values reported from countries where only the best measurement was provided were therefore corrected before analysis. Otherwise, it could have been the case that the reference curves overestimate the true status of the CF patients, as a consequence of analyzing mostly best measurements. However, additional analysis of the raw data, without corrections, yielded identical results, showing that the impact was overall small (see supplemental material). Thus the common assumption that selective reporting of best lung function can confound data interpretation is not supported by our findings.

One other result of this study is that the European CF BMI percentiles were in good agreement with the normative WHO 2007 curves up to age 7 but lower thereafter 16 . Despite recommendations to achieve greater fat and calorie intake, CF children and teenagers typically consume similar nutritional amounts as their healthy peers 23 24 . A positive association has been observed between a better nutritional status and a higher pulmonary function, with an inverse relation to morbidity and mortality 6 25 . The BMI of CF females remained closer to the WHO reference charts, while it has been reported that they experience steeper trajectories of health decline 26 27 .

In the inter-country comparison, we conjoined countries with a small number of patients and large coverage, and all countries with low coverage. This was done to limit the effect of chance variation that could arise from small populations. The outcomes in FEV₁ were somewhat larger in countries with small coverage. As it is not possible to rule out selection bias such as a survivor effect coupled to under diagnosis in these countries, this result should not be taken as evidence of better outcomes. The BMI inter-country comparison was less affected by coverage.

Although the FEV₁ CF specific percentiles provide a useful approach for comparing CF patients to their peers, they do not correct for attrition due to mortality 18 . For example, the population median CF percentile is 50% at all ages, but it is obvious that ranking 50^th among 8 years old CF patients is different from ranking 50^th at age 40. The use of “survival adjusted” CF specific percentiles has been proposed to measure severity as a quantitative trait irrespective of age 18 . More data will be required to fully adapt this method to the European situation which should become easier as neonatal screening takes hold across Europe thus significantly reducing ascertainment bias. In the meantime, the excellent correlation of the percentiles found from the European and the US analyses will in future allow cross comparison with US derived “common” phenotype in international studies involving North-America and Europe. This was indeed a critical limiting factor in earlier analyses 28 29 .

To conclude, although achieving normality is the ultimate goal of CF care, separately referencing against age and sex matched CF peers provides additional information to compare CF populations and better illustrates the range of variability between patients. These new reference equations also provide tools for computing quantitative traits for use in genome wide analyses 30 . Ours is only a first step towards the possibility to a fair comparison of European CF patients and health system performance. With the future availability of large phenotypic databases, it might be possible to apply our approach in other rare diseases, an emerging priority across the globe 31 .

The authors declare that have no competing interests.

PYB and HC designed the study and wrote the manuscript. PYB and PFB analysed the data. AM, LV and HO have been involved in the conception and design of the study, in the acquisition and interpretation of the data and in critically revisiting the manuscript. SR, MS, BA, CB, PD, MT, UK, MMZ, JFV and AC have been involved in the acquisition of the data and in revisiting the manuscript. All authors read and approved the final manuscript.